![List of Figures
3.1 Simplified view of the generic observer/controller architecture . . . . 14
4.1 Generic observer/controller architecture with two-level learning . . . . 28
4.2 Generic observer architecture consisting of a monitor, a pre-processor,
a data analyser, a predictor, and an aggregator . . . . . . . . . . . . . 29
4.3 Example of order perception: Depending on the objective of the
observer the nine balls are perceived as orderly or unorderly (position
on the left hand side vs. colour on the right hand side) . . . . . . . . 34
4.4 Fingerprint with different attributes at three specific times t0, t1, and
t2, visualised as a six-dimensional Kiviat graph (one dimension for
each attribute) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.5 Entropy values depending on the probability of the colour red . . . . 38
4.6 Generic controller architecture with two-level learning . . . . . . . . . 43
4.7 Centralised and distributed variants of the generic observer/controller
architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.8 Multi-levelled or hierarchical variant: An observer/controller on each
system element as well as one for the whole technical system . . . . . 51
4.9 Structure of an autonomic element, which interacts with other elements
and with human programmers via its autonomic manager, see [KC03] 53
4.10 Structure of the operator/controller module, see [HOG04] . . . . . . . 55
4.11 A mobile robot control system is decomposed traditionally into func-
tional modules, see [Bro86] . . . . . . . . . . . . . . . . . . . . . . . . 58
4.12 Task achieving behaviours as decomposition criterion for mobile robots,
see [Bro86] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.13 Control is layered in a hierarchy of levels of competence, where higher
layers subsume lower layers in the case of taking control, see [Bro86]:
Partitioning the system is possible at every level, the lower layers form
a complete operational control system . . . . . . . . . . . . . . . . . . 59
4.14 Three-levelled architecture for self-managed systems, see [KM07] . . . 62
XXI](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-27-320.jpg)
![List of Figures
5.1 The Woods101 example is a non-Markov environment . . . . . . . . . 77
5.2 Schematic overview of an XCS, see [Wil98] . . . . . . . . . . . . . . . 78
5.3 Variants of parallel LCSs as part of the single-agent learning approach:
Parallelism is distinguished on different levels, see [Gia97] . . . . . . . 86
5.4 Population structures for parallel multi-agent LCSs . . . . . . . . . . 89
5.5 Multi-agent learning approach . . . . . . . . . . . . . . . . . . . . . . 91
5.6 Two-level learning architecture is applied to an XCS . . . . . . . . . . 97
6.1 Variants of the predator/prey example, see [SV00] . . . . . . . . . . . 106
6.2 Snapshots of the chicken simulation: Unwounded chickens are white,
wounded chickens are dark (red), and feeding troughs are represented
by four bigger (yellow) circles. . . . . . . . . . . . . . . . . . . . . . . 110
6.3 An Eurovent cage with 60 chickens . . . . . . . . . . . . . . . . . . . 111
6.4 Finite state machine of a chicken representing the local behaviour
rules of a single chicken . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.5 Operational sequence of the chicken simulation, the contained observ-
ing and controlling steps are shown in Figures 6.7 and 6.12 . . . . . . 115
6.6 The generic architecture is applied to the chicken scenario . . . . . . 116
6.7 Steps of the observing process . . . . . . . . . . . . . . . . . . . . . . 117
6.8 Method to predict clustering, see [MMS06] . . . . . . . . . . . . . . . 119
6.9 Emergence value of the x-coordinates over time without any control
action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.10 Interpolated emergence value of the x-coordinates over time without
any control action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.11 Number of killed chickens #kc over time (every peak denotes a killed
chicken) without control action . . . . . . . . . . . . . . . . . . . . . 121
6.12 Steps of the controlling process . . . . . . . . . . . . . . . . . . . . . 121
6.13 Snapshot of the chicken simulation with noise control . . . . . . . . . 123
6.14 Controlling with fixed single rules . . . . . . . . . . . . . . . . . . . . 124
6.15 Emergence value of the x-coordinates over time with control action . 125
6.16 Interpolated emergence value of the x-coordinates over time with
control action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.17 Number of killed chickens #kc over time (every peak denotes a killed
chicken) with control action . . . . . . . . . . . . . . . . . . . . . . . 126
6.18 If a predefined threshold exceeds, learning will start using an XCS . . 126
6.19 Learning all possible situations using an XCS . . . . . . . . . . . . . 127
6.20 An XCS is equipped with a simulation model on level 2 . . . . . . . . 128
6.21 Simplified chicken scenario . . . . . . . . . . . . . . . . . . . . . . . . 130
6.22 Example with identical entropy and emergence values, respectively . . 131
6.23 An architectural overview of organic traffic control: Level 0 represents
the traffic node, levels 1 and 2 are organic control levels responsible
for the selection and generation of signal programmes, see [RPB+
06] . 132
XXII](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-28-320.jpg)
![List of Abbreviations
#kc – The number of killed chickens
2PXCS – Two parallel instances of the extended classifier system
3PXCS – Three parallel instances of the extended classifier system
d, dj – The duration of a noise signal
ex, ey, eh – The relative emergence indicators
HXCS – The hierarchical organised extended classifier system
i, ij – The intensity of a noise signal
LCS – A learning classifier system
L2 – Learning (or planning) on level 2 of the generic observer/controller archi-
tecture
OC – Organic computing
SuOC – The system under observation and control
tx, ty, th – Predefined thresholds of critical emergence values
XCS – The extended classifier system, as introduced in [Wil95]
XXV](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-31-320.jpg)
![Chapter1
Introduction
In the Nevada desert, an experiment has gone horribly wrong. A cloud
of nanoparticles – micro-robots – has escaped from the laboratory. This
cloud is self-sustaining and self-reproducing. It is intelligent and learns
from experience. For all practical purposes, it is alive. It has been
programmed as a predator. It is evolving swiftly, becoming more deadly
with each passing hour. Every attempt to destroy it has failed. And we
are the prey.
As fresh as today’s headlines, Michael Crichton’s most compelling novel
yet tells the story of a mechanical plague and the desperate efforts of a
handful of scientists to stop it. Drawing on up-to-the-minute scientific fact,
Prey takes us into the emerging realms of nanotechnology and artificial
distributed intelligence – in a story of breathtaking suspense. Prey is a
novel you can’t put down. Because time is running out. [Cri02]
These words cited above are written on the hardcover version of the techno-thriller
novel Prey by Michael Crichton. Of course, the book is science fiction and human
beings fighting against a swarm of micro-robots seems to be not realistic so far, but
an interesting story is told that features relatively new advances in computer science,
such as artificial life, swarm intelligence, self-organisation, genetic algorithms, or
multi-agent-based computing. Major themes of the book deal with the threat of
intelligent micro-robots escaping from human control and becoming autonomous,
self-replicating, and, by that, dangerous. Many aspects of the story, such as the
cloud-like nature of the nanoparticles and their nature-inspired process of evolution
closely follow research done by computer scientists in the past (few) years, see
e. g., the fields of evolutionary computation [FOW66, Gol89, Hol75], computational
intelligence [Eng02, PMG98], or artificial life1
.
1
http://www.alife.org](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-33-320.jpg)
![Chapter 1 Introduction
As short-sighted decision-making at the corporate level can lead to a disaster when
the companies involved control dangerous new technology, the book is about the
potential consequences, if suitable controls are not placed on biotechnology, before it
will develop to such an extent that it can threaten the survival of life on earth. Of
course, this is an important discussion and scientists doing research in informatics or
biologically inspired informatics have to cope with it.
Hence, this thesis will focus on the challenge of designing technical systems, which
are inspired by swarm intelligence, multi-agent systems, or self-organisation and
enable controllability of these systems at the same time. The research on these
different domains has intensified. A growing number of conferences2
, workshops3
,
and journals4
supports this trend.
Paradigms of computing are emerging based on modelling and developing compu-
ter-based systems exploiting ideas that are observed in nature. The human body’s
autonomic nervous system inspires the design of self-organising computer systems,
as proposed in IBM’s autonomic computing initiative5
. Some evolutionary systems
are modelled in analogy to colonies of ants or other insects. Highly-efficient and
highly-complex distributed systems are developed to perform certain functions or
tasks using the behaviour of insects as inspiration, e. g., swarms of bees, flocks of
birds, schools of fish, or herds of animals.
Self-organising systems are not science fiction any more, but problems with in-
creasing complexity and controllability of technical systems call for new system
architectures, as postulated in the field of organic computing (OC) and explicitly
investigated in this thesis.
1.1 Motivation
As mentioned in [BMMS+
06], the impressive progress in computing technology over
the past decades has not only led to an exponential increase in available computing
power, but also to a shrinking of computer chips to a miniature format. While only
twenty years ago, the predominant computing platform was a company mainframe
shared by many users, today, a multitude of embedded computing devices surrounds
us, including PDA, cell phone, digital camera, navigation system, MP3-player, etc. in
everyday life. An additional trend during recent years has been that these devices are
equipped with (often wireless) communication interfaces, allowing them to interact
and exchange information.
2
E. g., the International Joint Conference on Autonomous Agents and Multi-agent Systems
(AAMAS) or the Genetic and Evolutionary Computation Conference (GECCO)
3
E. g., the International Workshop on Learning and Adaptation in Multi-agent Systems (LAMAS)
or the International Workshop on Learning Classifier Systems (IWLCS)
4
E. g., Artificial Life (MIT Press Journals) or ACM Transactions on Autonomous and Adaptive
Systems (TAAS)
5
http://www.research.ibm.com/autonomic or see Section 4.5.1
2](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-34-320.jpg)
![1.1 Motivation
Amongst others this outlook towards smaller, more intelligent, and more numer-
ous devices surrounding everybody in his everyday life is given by the paradigm
of ubiquitous computing, which was first introduced by Mark Weiser in [Wei91].
Future information processing will be integrated into a broad range of everyday
and everywhere objects making these objects intelligent. These devices will be
interconnected and they will communicate over various communication channels.
Thus, networks of intelligent systems will grow, and their behaviour will no longer
be predictable with certainty due to interaction effects, see [Sch05a].
In addition, other large technical systems consist of more and more interconnected
electronic devices. For example, in cars, numerous processors and embedded systems
keep the vehicle on the road, control the engine with respect to combustion and
pollution, assist the driver, provide security with air bags and seat belt systems,
provide functions such as air conditioning, navigation, parking assistant, information
services, and entertain the passengers. All these controllers are connected to a
complex communication network. And this development has not stopped yet.
Technical innovations are only a stone’s throw away from scenarios like smart
factory, with flexible robots self-organising to satisfy the needs at hand [Gui08], or
smart cars that adapt to different drivers and road conditions, communicate with
other cars on special events, or integrate personal devices (PDA, mobile telephone,
or notebook) into their network.
While this development is exciting, the resulting systems become increasingly
complex, up to the point where they can no longer be designed or used easily. Even
today, in the automotive sector, it is estimated that about half of all car break-downs
are caused by electric and electronic components. E. g., in 2005 weak car batteries
head the table of causes for break-downs listed by the ACE Auto Club Europa with
25% [ACE06], while electronic components are listed on third position (currently)
with 13%, still increasing their percentage.
Thus, the questions arise, how to design such complex distributed and highly
interconnected systems, and how to make them reliable and usable. Clearly, the
designer is not able to foresee all possible system configurations, and to prescribe
proper behaviours for all cases. Additionally, the user is relieved from having control
in detail over all parameters of the system, allowing him or her to influence the
system on a higher level, e. g., by setting goals.
OC has the vision to meet this challenge of coping with increasing complexity by
making technical systems more life-like, and endowing them with properties such
as self-organisation, self-configuration, self-repair, or adaptation [Sch05b]. Future
systems will possess certain degrees of freedom to handle unforeseen situations and
act in a robust, flexible, and independent way. Thus, these systems will exhibit
self-organised behaviour, which makes them able to adapt to a changing surrounding.
That is why, in the field of OC, these systems are called organic. Hence, an OC system
is a system, which dynamically adapts to the current situation of its environment, but
still obeys goals set by humans. In addition to this environmental awareness, systems
3](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-35-320.jpg)
![Chapter 1 Introduction
providing services for humans will adjust themselves to the users’ requirements (and
not vice versa). Based on these trends, the question, also addressed in this thesis,
is not, whether complexity increases or informatics is confronted with emergent
behaviour, but how new technical systems will be designed that have the possibility
to cope with the emerging global behaviour of self-organising systems by adequate
control actions.
1.2 Objectives and Approach
As outlined before and motivated in Chapter 2, OC has a major research interest in
new system architectures that self-organise and adapt exploiting certain degrees of
freedom. To achieve these ambitious goals of designing and controlling self-organising
systems, adequate methods and system architectures have to be developed, since no
general approach exists to build OC systems. Therefore, OC proposes a regulatory
feedback mechanism, the so-called generic observer/controller architecture [MS04],
which constitutes one way to achieve controlled self-organisation in technical systems.
Using this control loop, an organic system will adapt over time to its changing
environment. It is obvious that this architecture could benefit from learning capabil-
ities to tackle these challenges. Therefore and as described in detail in Chapter 4,
the controller has been refined by a two-levelled learning approach. At the on-line
level (level 1) the proposed architecture learns about the environment, and about
the performance of its control strategies. Level 2 implements a planning capability
based on a simulated model of the environment. At this level an agent can test and
compare different alternative strategies off-line, and plan its next action without
actually acting in the environment. Thus, the two research questions addressed in
this thesis are defined in the following in more detail.
1. What does it mean to establish and utilise controlled self-organisation in the
context of technical OC scenarios specially focussing on learning classifier
systems (LCSs) as machine learning technique on level 1 of the proposed
two-levelled learning architecture?
2. How is the (on-line) learning process speeded up?
In other words, the observer/controller architecture is refined into a form that
can serve as a generic template containing a range of components, which should be
necessary in a range of OC application scenarios. It enables a regulatory feedback
mechanism and the use of machine learning techniques to improve a single-agent’s or a
multi-agent system’s behaviour in technical domains with the following characteristics.
• There exists a need for on-line decision-making,
• decisions are based on (aggregated) sensor information,
4](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-36-320.jpg)
![Chapter 1 Introduction
While LCSs have drawbacks in learning speed, which seem to be critical in
combination with technical applications, mechanisms have been investigated, which
speed up the learning process of LCSs. These approaches are compared with the
proposed two-levelled learning architecture that learns on-line (level 1) about the
environment and about the performance of its control strategies, while on level 2 a
planning capability is used, based on a simulation model of the environment, where
an agent can test and compare different alternative strategies off-line, and thus plan
its next action without actually acting in the environment.
1.3 Major Contributions
In brief, this thesis makes three main contributions related to the research fields of
OC. First, OC research is summarised that has been done over the last five years
and specially focusses on the design of the generic observer/controller architecture,
which serves as a framework for building OC systems. This architecture allows
for self-organisation, but at the same time enables adequate reactions to control
the – sometimes completely unexpected – emerging global behaviour of these self-
organised technical systems. The proposed architecture can be used in a centralised,
distributed, or multi-levelled and hierarchically structured way to achieve controlled
self-organisation as a new design paradigm. Thus, Chapters 3 and 4 address related
work in the field of architectures for controlled self-organisation. Some contents
have been published in [ÇMMS+
07, SMSÇ+
07, SMS08]. Chapter 4 mainly bases on
[BMMS+
06, RMB+
06].
Secondly, the idea of a two-level learning approach is introduced as part of the
controller. Since a learning capability is an essential feature of OC systems, the generic
architecture and, in particular, the controller, has to include adequate components
for learning. The work, presented in this thesis, focusses on on-line learning and
specially on the investigation of LCSs as an adequate machine learning technique.
Thirdly, several distributed variants of LCSs are investigated with the objective of
improving learning speed and effectiveness. While conventional LCSs have drawbacks
in learning speed, this thesis investigates possible modifications by decomposing a
problem into smaller sub-problems and by learning these subtasks independently.
Furthermore, the performance of these variants of on-line LCSs are compared to the
combination of on-line learning and off-line planning capabilities, as suggested by
the observer/controller architecture.
Since the second and the third contributions are inherently domain-specific, the
following chapters provide a general specification as well as an implementation within
multi-agent test scenarios. The used multi-agent test scenarios are selected from the
predator/prey domain and can therefore be generalised to other domains. Also, work,
which has been done in [RRS08], is shortly summarised in Chapter 6. Chapters 5,
6, and 8 present results that have been published in [MRB+
07, RM08, RPS08].
6](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-38-320.jpg)
![Chapter 2
Organic Computing (OC)
It is not the question, whether adaptive and self-organising systems will
emerge, but how they will be designed and controlled. [Sch05b, SMSÇ+
07]
Since the work presented here is mainly based on OC research topics, this chapter
summarises the vision of OC. Motivation and challenges of this young research field
are explained in short, before the contributions of this thesis are described in the
following chapters.
As outlined in Section 1.1, the increasing complexity of technical systems calls
for research into new design principles. It is impossible for a designer to foresee all
possible configurations and to explicitly specify the entire behaviour of a complex
system on a detailed level. In particular, if the system consists of many interacting
components, it may exhibit new, emergent properties that are very difficult to
anticipate. Emergent phenomena are often identified, when the global behaviour
of a system appears more coherent and directed than the behaviour of individual
parts of the system (the whole is more than the sum of its parts). More generally,
such phenomena arise in the study of complex systems, where many parts interact
with each other and where the study of the behaviour of individual parts reveals
little about system-wide behaviour. Especially, in the area of multi-agent systems
emergence and self-organisation have been studied extensively, see [DFH+
04, DGK06]
for two recent surveys.
Despite their complexity, living creatures are very robust and have the natural
ability to learn and adapt to an uncertain and dynamic environment. The idea of
OC is therefore to address complexity by making technical systems more life-like
and to develop an alternative to the explicit total a priori specification of a system.
Instead, organic systems should adapt and self-organise with respect to some degrees
of freedom. But, OC systems should be designed with respect to human needs, and
have to be trustworthy, robust, adaptive, and flexible. They will show the so-called
self-x-properties: Self-configuration, self-optimisation, self-healing, self-explanation,](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-41-320.jpg)
![Chapter 2 Organic Computing (OC)
and self-protection. Such systems are expected to learn about their environment
during life time, will survive attacks and other unexpected breakdowns, will adapt to
their users, and will react sensibly, even if they encounter a new situation, for which
they have not been programmed explicitly. In other words, an OC system should
behave more life-like (organic).
This can only be achieved by adding some kind of awareness of their current
situation to the system elements and the ability to provide appropriate responses to
dynamically changing environmental conditions. The principles of OC are strongly
related to the objectives of IBM’s autonomic computing initiative, see Section 4.5.1.
But, while autonomic computing is directed towards maintaining server architectures,
which should be managed without active interaction between man and machine
[KC03, Ste05], OC’s focus is more general in its approach and addresses large
collections of intelligent devices, providing services to humans, adapted to the
requirements of their execution environment [Sch05b]. Thus, besides showing the
self-x-properties, interaction between man and machine is an essential part of OC
systems.
The term organic computing was formed in 2002 as a result of a workshop aiming at
future technologies in the field of computer engineering. The outlines of the workshop
and the OC vision were first formulated in the joint position paper [ACE+
03] of
the section of computer engineering (Technische Informatik) of the Gesellschaft für
Informatik (German Association for Informatics, GI) and the Informationstechnische
Gesellschaft (German Association for Information Technology). In 2005, the German
Research Foundation (Deutsche Forschungsgemeinschaft, DFG) approved a priority
research programme on OC for six years (2005–2011). This research programme
addresses fundamental challenges in the design of OC systems; its goal is a deeper
understanding of emergent global behaviour in self-organising systems and the
design of specific concepts and tools to construct and control OC systems for
technical applications. Topics, such as adaptivity, reconfigurability, emergence of
new properties, and self-organisation, play a major role. Currently, the research
programme provides funding for 18 research projects with a total volume of around
EUR 2 million per year. The topics of these projects range from traffic control over
robot coordination to chip design. Information on the different projects can be found
via the OC website1
.
Self-organising systems bear several advantages compared to classical, centrally
controlled systems. Amongst others, the failure of a single component should not
cause a global malfunction of the whole system. Such a system will be able to adapt
to changing circumstances. As a result, self-organisation could be described as a
method of reducing the complexity of computer systems.
In such self-organising systems the local interaction of the system elements may
result in an emergent global behaviour, which can have positive (desired) as well as
1
http://www.organic-computing.de/spp
10](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-42-320.jpg)
![negative (undesired) effects. Self-organisation and emergent phenomena also initiate
new problems unknown in the engineering of classical technical systems. A global
emergent behaviour usually is a nonlinear combination of local behaviours. Its design
process with both potential design directions (top-down vs. bottom-up design) turns
out to be a highly non-trivial task: For a top-down approach it is hard to deduce
adequate local rules from a desired global behaviour, and in the bottom-up direction
it quite often remains unclear how local rules go together with global behaviour,
see [KC03].
In this context and in order to assess the behaviour of the technical system and
– if necessary – for a regulatory feedback to control its dynamics, the so-called
observer/controller architecture has become widespread in the OC community as a
design paradigm to assure the fulfilment of system goals (given by the developer or
user), see Chapter 4. The observer/controller uses a set of sensors and actuators to
measure system variables and to influence the system. Together with the system
under observation and control (SuOC), the observer/controller forms the so-called
organic system. An observer/controller loop enables adequate reactions to control the
– sometimes completely unexpected – undesired emerging global behaviour resulting
from local agents’ behaviour.
However, besides this fascinating outlook, the materialisation of the OC vision
depends on several crucial factors, which are summarised in [Sch05b].
• Designers of OC systems have to guarantee that self-organising systems, based
on OC principles, do not show unwanted (emergent) behaviour. This is
particularly important, when malfunction can have disastrous consequences,
e. g., in safety critical applications. The generic observer/controller architecture,
as described in Chapter 4, seems to be a promising approach in asserting certain
functionality and additionally in keeping the system at an effective state of
operation. OC systems will only become accepted, if users can trust them.
Therefore, trust/reliability could turn out to be the most important prerequisite
for acceptance.
• Closely related is the need for the user to monitor and influence the system: It
has to be guaranteed that it is still the user, who guides the overall system.
Therefore, the system developer has to design user interfaces, which can be used
to control the system – which means that there has to be a possibility to take
corrective actions from outside the system. The generic observer/controller
architecture considers this requirement.
• Developers of OC systems have to determine appropriate rules and patterns
for local behaviour in large networks of smart devices in order to provide some
requested higher functionality. Important topics in the design of self-organising
systems are the utilisation of arising emergent phenomena and controlling the
local level in such a way that the system shows the desired behaviour at global
11](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-43-320.jpg)
![Chapter 3
Controlled Self-Organisation
Technological systems become organised by commands from outside, as
when human intentions lead to the building of structures or machines. But,
many natural systems become structured by their own internal processes:
These are self-organising systems, and the emergence of order within them
is a complex phenomenon that intrigues scientists from all disciplines.
[YGWY88]
Self-organising systems are well known from nature and have been studied in
domains like physics, chemistry, and biology. In recent years research interest has
succeeded to apply concepts of self-organisation to technical systems. The reason is
a paradigm shift from monolithic systems to large networked systems driven by the
technological change of integrating more information processing into the everyday
life, objects, and activities. The necessity to find new approaches to cope with
upcoming problems of increasing complexity attracts awareness to the principle of
self-organisation.
OC systems should use self-organisation to achieve a certain externally provided goal.
Furthermore, the system has to adapt to changing environmental requirements and to
be capable to deal with (unanticipated) undesired emergent behaviour. Therefore, OC
systems are assumed to support controlled self-organisation. Whenever necessary, this
requires a range of methods for monitoring and analysing the system performance and
for providing appropriate control actions. The generic observer/controller architecture
– this architecture is introduced in detail in Chapter 4 – promises to provide the
necessary components for satisfying all these demands, see Figure 3.1.
Similar to the MAPE cycle (monitor, analyse, plan, and execute) of IBM’s au-
tonomic computing initiative, see Section 4.5.1, a closed control loop is defined to
keep the properties of the self-organising SuOC within preferred boundaries. The
observer observes certain (raw) attributes of the system and aggregates them to
situation parameters, which concisely characterise the observed situation from a](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-45-320.jpg)
![Chapter 3 Controlled Self-Organisation
organic system
SuOC
goals
: agent/robot/entity
system status
observer controller
observes
controls
reports
input output
learning
Figure 3.1: Simplified view of the generic observer/controller architecture
global point of view, and passes them to the controller. The controller acts according
to an evaluation of the observation (which might include the prediction of future
behaviour). If the current situation does not satisfy the requirements, it will take
action(s) to direct the system back into its desired range, will observe the effect of
the intervention(s), and will take further actions, if necessary. Using this control loop
an organic system will adapt over time to its changing environment. It is obvious
that the controller could benefit from learning capabilities to tackle these challenges.
Although the observing and controlling process is executed in a continuous loop, and
the SuOC is assumed to run autonomously, even if the observer/controller archi-
tecture is not present – even though in a suboptimal way. Furthermore, emergence
plays a central role in OC systems. Emergent and self-organising behaviour has been
observed in nature, demonstrated in a variety of computer simulated systems in
artificial life research, and it has also occurred in highly complex technical systems,
where it has quite often led to unexpected global functionality [BDT99]. Despite the
importance of a rigorous description of these phenomena, the quantitative analysis
of technical self-organising systems is still a rather unexplored area. Therefore, this
chapter describes the understanding of the basic mechanisms of self-organisation and
emergent behaviour in complex (organic) ensembles, summarises related work, and
provides appropriate (metrics and) tools for utilising controlled self-organisation.
In Section 3.1 research is summarised that has been done in the area of self-or-
ganisation and in Section 3.2 about the related concept of emergence. There may
be instances of self-organisation without emergence and emergence without self-
organisation, and there is evidence in literature that the phenomena are not the same.
14](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-46-320.jpg)
![3.1 Self-Organisation
However, future research is needed to clarify the relation between these two terms.
Finally, in Section 3.3 an architectural-based approach for the design and engineering
of technical systems is proposed that makes use of controlled self-organisation, before
describing the OC approach in Chapter 4.
3.1 Self-Organisation
The dynamics of a system can tend by themselves to increase the inherent order
of a system. This idea has a long history, being first introduced by the French
philosopher René Descartes. In 1947, the term self-organisation was introduced
by the psychiatrist and engineer William Ross Ashby [Ash47]. Cyberneticans, like
Heinz von Foerster, Stafford Beer, Gordon Pask, and Norbert Wiener, took up this
concept and associated it with general systems theory in the 1960ies. In the 1970ies
and 1980ies physicists adopted self-organisation to the field of complex systems and
established the topic in scientific literature. Even if the concept of self-organisation
is very promising to solve complex problems, as explained in [Ger07], the notion of
self-organisation will remain somewhat vague, and discussion has been widespread.
The extensive FAQ-list1
is a good link to research that has been done so far.
The term self-organisation is used frequently, but a generally accepted meaning has
not emerged. As the list grows, it becomes increasingly difficult to decide whether
these phenomena are all based on the same process, or whether the same label
has been applied to several different processes. Despite its intuitive simplicity as
a concept, self-organisation has proven notoriously difficult to describe and define
formally or mathematically. Thus, it is entirely possible that any precise definition
might not include all the phenomena, to which the term of self-organisation has been
applied. In the following, it will not be attempted to give a new definition facing
the philosophical problem of defining self, the cybernetic problem of defining system,
or the universal problem of defining organisation. Instead, research is summarised
that has been done so far to characterise the conditions necessary to call a system
self-organising. Answers will be given to the following questions: What is a self-
organising system? What is it not? And what are possible approaches to engineer
self-organising technical systems?
3.1.1 Understanding Self-Organisation from the Viewpoint of Different
Sciences
As pointed out by Carlos Gershenson in [Ger07], the term self-organisation has been
used with different meanings, e. g., in computer science [HG03, MMTZ06], biology
[CDF+
03, FCG06], mathematics [Len64], cybernetics [Ash62, von60], synergetics
1
http://www.calresco.org/sos/sosfaq.htm
15](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-47-320.jpg)
![Chapter 3 Controlled Self-Organisation
[Hak81], thermodynamics [NP77], complexity [Sch03], information theory [Sha01],
and evolution of language [de 99]. Selected ideas, which specially contribute to the
idea of OC, are summarised from the viewpoint of different sciences.
Self-Organisation in Nature
According to [CDF+
03], self-organisation in biological systems is often described as
a process, in which a pattern at the global level of a system solely
emerges from numerous interactions among the lower level components
of the system. Moreover, the rules specifying interactions among the
system’s components are executed using only local information, without
reference to the global pattern [YGWY88].
A self-organising system in nature acts without centralised control and operates
according to local contextual information. Thus, spontaneous behaviour without
external control produces a new organisation reacting to environmental changes/dis-
turbances. Natural systems often show a robust behaviour, they adapt to changes,
and they are able to ensure their own survivability. There are quite a few examples
of natural systems, which are not at all robust (in particular at the individual
level) and which cannot adapt to changes. Robustness often is a property of a
population/swarm, not of an individual. In some cases, self-organisation is linked
to emergent behaviour, as described later in Section 3.2. Individual components
carry out a simple task, and as a whole these components are able to carry out a
complex task emerging in a coherent way through the local interactions of various
components. Typical examples from nature are found in the following.
• Social insects, like ants, termites, or honey bees, where communication oc-
curs through stigmergy by placing chemical substances, pheromones, into the
environment. In 1959 the theory of stigmergy has been defined as the work
excites the workers by [Gra59]. Direct interactions between the individuals
are not necessary to coordinate a group. Indirect communications between
the individuals and the environment are enough to create structures. Thus,
coordination or regulation tasks are achieved without centralised control.
• Flocks of birds or schools of fish, where collective behaviour is defined by simple
rules, like getting close to a similar bird (or fish) – but not too much – and
getting away from dissimilar birds (or fishes) to collectively avoid predators.
• Social behaviours of humans, where emergent complex global societies arise by
working with local information and local direct or indirect interactions.
• The immune system of mammalians, where cells regenerate self-organised.
16](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-48-320.jpg)
![3.1 Self-Organisation
From a very general point of view, the notion of autopoiesis is often associated
with the term of self-organisation. In the 1970ies biological studies established
autopoiesis (meaning self-production) [MV91, Var79], which describes the process
of a living system as an organisation to produce itself. An autopoietic system is
autonomous and operationally closed, in the sense that every process within it
directly helps maintaining the whole. For example, cells or organisms self-maintain
the system through generating system’s components. The cell is made of various
biochemical components such as nucleic acids and proteins, and is organised into
bounded structures such as the cell nucleus, various organelles, a cell membrane, and
cytoskeleton. These structures, based on an external flow of molecules and energy,
produce the components, which, in turn, continue to maintain the organised bounded
structure that gives rise to these components.
In computing, an analogous concept is called bootstrapping, which refers to tech-
niques that allow a simple system to activate a more complicated system, e. g.,
when starting a computer a small programme, the built in operating system (BIOS),
initialises and tests the hardware (peripherals or memory), loads another programme,
and passes control to this programme (an operating system).
Self-Organisation in Chemistry
Self-organisation is also relevant in chemistry, where it has often been taken as
being synonymous with self-assembly. To name a few, this includes research in
molecular self-assembly, which is the process, by which molecules adopt a defined
arrangement without guidance or management from an outside source [Leh88, Leh90].
Additionally, self-organisation is used in the context of reaction diffusion systems,
which are mathematical models that describe, how the concentration of one or more
substances is distributed in space changes. This occurs under the influence of two
processes that are local chemical reactions, in which the substances are converted
into each other, and diffusion, which causes the substances to spread out in space
[Fif79]. Other examples are autocatalytic networks, or liquid crystals.
Through thermodynamics studies [GP71] the term self-organisation itself has been
established in the domain of chemistry in the 1970ies. When an external energy
source is applied to an open system, this system decreases its entropy (where order
comes out of disorder, see Section 4.1.5). In other words, a system reaches a new
system state, where entropy is decreased, when external pressure is added. Compared
to the stigmergy concept, mentioned in Section 3.1.1, where self-organisation results
from a behaviour occurring inside the system (from the social insects themselves
placing pheromones in the environment), this is a fundamental difference from the
understanding in chemistry. In the latter case, self-organisation seems to be a result
of an external pressure applied from the outside.
17](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-49-320.jpg)
![Chapter 3 Controlled Self-Organisation
Self-Organisation in Mathematics and Computer Science
Self-organisation has also been observed in mathematical systems such as cellular
automata [TGD04]. In computer science, some instances of evolutionary computation
and artificial life exhibit features of self-organisation.
Research in artificial systems has been oriented towards introducing self-organi-
sation mechanisms specifically for software applications, see [BDHZ06, BDKN05,
BHJY07, DKRZ04]. These applications have been inspired by already mentioned
nature-inspired concepts like stigmergy, autopoiesis, or the holon concept introduced
by Artúr Kösztler in [Kös90]. The term holon describes systems, which represent
whole systems and parts of larger systems at the same time. Then, holarchies describe
hierarchies of such holons. Typical examples of self-organising artificial systems are
swarm-inspired techniques for routing [BDT99] or load-balancing [MMB02].
Furthermore in multi-agent systems, (software) agents play the role of self-
organising autonomous entities. Frequently, multi-agent systems are used for simulat-
ing self-organising systems, in order to get a better understanding of the dependencies
in such systems or to establish models of the simulated systems. As mentioned in
[DGK06], the tendency of initiatives like OC or autonomic computing is now to
shift the role of agents from simulation to the development of distributed systems.
Components (e. g., software agents) that once deployed self-organise in a predefined
environment and work in a distributed manner towards the realisation of a given
(global) possibly emergent functionality.
3.1.2 Properties of Self-Organisation
In general, the understanding of self-organisation seems to be widespread. According
to [DGK06] self-organisation essentially refers to a spontaneous and dynamically
produced (re-)organisation. Several, more qualitative, properties/issues should be
extracted from the different viewpoints mentioned above in the following section to
end up in a possible definition in Section 3.1.3.
Properties of Self-Organisation in Nature
According to swarm intelligence [BDT99], self-organising processes are characterised
by four properties.
1. Multiple interactions among the individuals,
2. retroactive positive feedback (e. g., increase of pheromone, when food is de-
tected),
3. retroactive negative feedback (e. g., pheromone evaporation), and
18](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-50-320.jpg)
![3.1 Self-Organisation
4. increase of behaviour modification (e. g., increase of pheromone, when new
path is found).
From the more biologically-inspired viewpoint of autopoiesis, a self-organising
system could be characterised as an autopoietic machine, which is a machine that is
organised (defined as a unity) as a network of processes of production
(transformation and destruction) of components, which
1. through their interactions and transformations continuously regen-
erate and realise the network of processes (relations) that produced
them; and
2. constitute it (the machine) as a concrete unity in space, in which
they (the components) exist by specifying the topological domain of
its realisation as such a network [MV91].
Properties of Self-Organisation in Chemistry
Under external pressure, self-organising behaviour is characterised by a decrease of
entropy and satisfies the following requirements, as stated in [GP71].
1. Mutual causality: At least two components of the system have a
circular relationship, each influencing the other.
2. Autocatalysis: At least one of the components is causally influenced
by another component, resulting in its own increase.
3. Far from equilibrium condition: The system imports a large amount
of energy from outside the system, uses the energy to help renew its
own structures (autopoiesis), and dissipates rather than accumulates,
the accruing disorder (entropy) back into the environment.
4. Morphogenetic changes: At least one of the components of the system
[has to] be open to external random variations from outside the sys-
tem. A system exhibits morphogenetic change when the components
of the system are changed themselves.
Properties of Self-Organisation in Artificial Systems
In [DGK04], two definitions of self-organisation in artificial systems have been
established. Self-organisation implies organisation, which in turn implies some
ordered structure as a result of component behaviour. A new distinct organisation is
self-produced, since the process of self-organisation changes the respective structure
and behaviour of a system.
19](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-51-320.jpg)
![Chapter 3 Controlled Self-Organisation
• Strongly self-organising systems are systems that change their or-
ganisation without any explicit, internal or external, central control.
• Weakly self-organising systems are systems where reorganisation
occurs as a result of internal central control or planning.
3.1.3 Definition of Self-Organisation
The previous sections have shown that it is not trivial to give a precise definition of
self-organisation. However, a practical notion, as given in [Ger07], will suffice for the
purposes of this thesis.
A system described as self-organising is one, in which elements interact
in order to dynamically achieve a global function or behaviour.
Furthermore, this self-organising behaviour is autonomously achieved through
distributed interactions between the system components, which produce feedbacks
that regulate the system. Instead of answering the question, which are the necessary
conditions for a self-organising system, another question can be formulated: When
is it useful to describe a system as self-organising? In [Ger07], it is argued that
self-organising systems (in the sense of distributed systems) will have advantages in
dynamic and unpredictable environments, where problems have to be solved that
are not known beforehand and/or the addressed problem changes constantly. Then,
a solution dynamically arises by local interactions and adaptation to unforeseen
disturbances quickly appears. In theory, a centralised approach is also able to solve
the problem, but in practice such an approach may require too much computation
time to cope with the unpredictable disturbances in the system and its environment,
e. g., when a system or its environment changes in less time than the system requires
to compute a solution.
In [Ed08], another brief sentence has been worked out to explain the main idea
of self-organisation, being similar to the definition given by Carlos Gershenson in
[Ger07].
A self-organising system consists of a set of entities that obtains an
emerging global system behaviour via local interactions without centralised
control.
3.1.4 Summary
The investigation of self-organisation in many different disciplines of science has
advantages and disadvantages at the same time. Many definitions from different
domains have blurred the whole idea, which definitely is a disadvantage in terms
of definition and terminology. On the other hand, the many disciplines keep the
20](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-52-320.jpg)
![3.2 Emergence
potential for many ideas and new approaches for creating (controlled) self-organising
systems. This possibility will be even more attractive, if the research on self-organising
systems converges towards a more standardised nomenclature, probably even forming
a new field of science some day.
Several positive effects from the interdisciplinarity of self-organisation became
apparent when discussing possible ways to design the behaviour of the particular
entities that form a self-organising system. The local behaviour is an integral part
of a self-organising system, since the whole behaviour of the system emerges from
the local interactions of the entities. In [Ed08], three basic approaches have been
identified for finding a suitable set of local rules: Nature-inspired design, trial and
error, and learning from an omniscient solution.
In the case of dynamic self-organising systems, with distributed control and local
interactions, emergence appears to be some kind of structure on a higher level. Since
literature investigating self-organisation is also linked to the term of emergence, this
phenomenon is addressed in the next section.
3.2 Emergence
The phenomenon of emergence has been a fascinating topic for scientists such as
John Stuart Mill [Mil43], George Henry Lewes [Lew75], and Conwy Lloyd Morgan
[Llo23] for a long time, and the philosophical discussion of this topic is more than
150 years old. These so-called proto-emergentists consider the emergent process as a
black box, where only the inputs and the outputs at the lowest level can be discerned
without any knowledge about how the inputs are transformed into outputs.
However, in the case of designing technical OC systems more recently characterised
aspects of emergence need to be considered. A different perspective, referred to as
neo-emergentism, summarises approaches of Jochen Fromm [Fro04, Fro05], John H.
Holland [Hol98], Stuart Kauffman [Kau93], Aleš Kubík [Kub03], and others, where
the root of emergence bases on the dynamics of a system, where investigations
focus on reproducing the process, which leads to emergence, and where emergent
phenomena are less miraculous than in the black box view.
Emergence is the phenomenon occurring when a population of interconnected
relatively simple entities self-organises to form more ordered higher level behaviour
[Joh01]. Emergence can be referred to as the effect that the whole is greater than the
sum of its parts. Emergent phenomena are defined by
1. the interaction of mostly large numbers of individuals
2. without centralised control with the result of
3. a global system behaviour, which has not explicitly been programmed into the
individuals [Bea03].
21](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-53-320.jpg)
![Chapter 3 Controlled Self-Organisation
The journal Emergence2
, a journal of complexity issues in organisation and man-
agement, provides the following characterisation of emergent behaviour.
The idea of emergence is used to indicate the developing of patterns,
structures, or properties that do not adequately seem explained by referring
only to the system’s pre-existing components and their interaction. Emer-
gence becomes of increasing importance, when the system is characterised
by the following features.
• When the organisation of the the system, i. e., its global order,
appears to be more salient and of a different kind than the components
alone;
• when the components can be replaced without an accompanying
decommissioning of the whole system;
• when the new global patterns or properties are radically novel with
respect to the pre-existing components; thus, the emergent patterns
seem to be unpredictable and non-deducible from the components as
well as irreducible to those components.
Good examples for emergence originate from the observation of ants and other
insects. The social insect metaphor for solving problems has become a diverse
topic during the last years [BDT99]. For example, foraging behaviour in ants is
characterised by the distribution of pheromones, thereby encouraging (but not
forcing) other ants to follow the paths. This behaviour, despite its simplicity and
distributedness, results in a very robust and efficient emergent phenomenon, i. e.,
that ants collectively find the shortest path between nest and food source. This
observation has resulted in powerful metaheuristics for solving complex problems,
called ant colony optimisation.
Another example for emergent behaviour is the (human) brain. Although the exact
function and interrelation of the different brain sub-systems is not really understood,
scientists assume underlying emergent effects, as explained in [Rot05].
Today’s neurobiology is able to investigate those processes in human and
animal brains in detail, which are responsible for the higher level cognitive
functions like object recognition, attention, memorising, thinking, problem
solving, action planning, empathy, and self-reflection, i. e., processes
usually related to consciousness. It shows that these functions can uniquely
be mapped to certain brain regions, and vice versa. This does not mean
a violation of known physical/chemical/physiological laws. Neither are
there any unexplainable gaps. Therefore, it seems necessary to view these
brain functions as emergent states of a physical system. This is not in
2
http://www.emergence.org
22](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-54-320.jpg)
![introduction of nude figures and elaborate landscape and architectural
settings. But in trying to copy painting too closely, they departed from
the best traditions of tapestry technique, and deterioration was sure to
follow in time.
After the desolating wars of the sixteenth century, when arts and
industries revived under the Archdukes Albert and Isabella, Brussels
weavers set up their looms again, and “Rubens brought new life into
tapestry manufacture. He supplied the Brussels workshops with four
great series—the History of Decius Mus, destined for some Genoese
merchants; the Triumphs and Types of the Eucharist, ordered by the
Infanta Isabella for the convent of the Clares at Madrid; the History of
the Emperor Constantine, executed for Louis XIII; and the History of
Achilles, for Charles I.... The Triumphs and Types of the Eucharist are
the most powerful allegories ever created to glorify the mysteries of
the Catholic religion.”[4]
Jacob Jordaens also designed tapestry cartoons, but the most popular
artist among the weavers at the end of the seventeenth and in the
eighteenth centuries was David Teniers. He did not himself make
designs, but the manufacturers, especially at Oudenarde, borrowed his
subjects, which were drawn largely from peasant and village life.
One reason why we have so few of the really antique tapestries is that
in 1797 the market for them was so dead—owing to the increasing use
of wall-papers and canvases painted in oils—that the French decided it
would be better to burn them for the gold and silver they contained.
Accordingly, “One hundred and ninety were burned. During the French
Revolution, a number of tapestries that bore feudal emblems were
also burned at the foot of the Tree of Liberty.” At this time, when they
were not in fashion, many rare old hangings were cut up by the
inartistic or the ignorant and used as rugs and curtains.
But in recent years, we are told, the Brothers Braquenié have set up a
workshop at Malines, where they have produced a fine series for the
Hôtel de Ville in Brussels, called “Les Serments et les Métiers de
Bruxelles.” The cartoons for this set were made by Willem Geefs, the
painter.](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-73-320.jpg)
![As to the material, there is a great difference. Gothic tapestries are
composed of woolen weft on linen, or woolen on hemp warp, and are
often enriched with gold and silver thread. These are not used today,
as they are considered too expensive. Since the sixteenth century,
Brussels, Gobelins, and Mortlake have used a great deal of silk. In the
fifteenth century fifteen or twenty colours were employed, in the
Renaissance period, twenty or thirty.
“Both high warp and low warp antedated the shuttle. In other words,
they use bobbins that travel only part way across instead of shuttles
that travel all the way across.” The high warp loom was also in use
before the treadle. “In the low warp loom the odd threads of the warp
are attached to a treadle worked with the left foot, the even threads
of the warp to a treadle worked with the right foot, thus making
possible the manipulation of the warp with the feet and leaving both
hands free to pass the bobbins. In the high warp loom, that has no
treadle, the warps are manipulated with the left hand while the right
hand passes the bobbins back and forth. The term high warp means
that the warp is strung vertically, low warp horizontally.”
Both are woven with the wrong side toward the weaver. “The wrong
side in all real tapestries is just the same as the right side except for
reversal of direction and for the loose threads.... In the high warp
loom, the outline of the design is traced on the warp threads with
India ink from tracing paper, and the coloured cartoon hangs behind
the weaver, where he consults it constantly. In the low warp loom, the
coloured cartoon is usually beneath the warp, and often rolls up with
the tapestry as it is completed.”[5] In the eighteenth century, the low
warp loom was considered better than the haute lisse, or high warp.
Great care has to be taken in dyeing the threads of the weft, which
are much finer than those of the warp. Vegetable dyes, such as
cochineal, madder, indigo, etc., must be used, for permanent colours
can never be obtained with aniline dyes. The old Spanish dyes were
considered the best. In this country, one sometimes gets the fine
colours in an old Mexican serape or a prized Navajo blanket. The wool
that is used to mend old tapestries in the American museums is](https://image.slidesharecdn.com/2321189-250519063841-3e9b1233/85/Controlled-Selforganisation-Using-Learning-Classifier-Systems-Urban-Maximilian-Richter-74-320.jpg)
Controlled Selforganisation Using Learning Classifier Systems Urban Maximilian Richter
- 1. Controlled Selforganisation Using Learning Classifier Systems Urban Maximilian Richter download https://ebookbell.com/product/controlled-selforganisation-using- learning-classifier-systems-urban-maximilian-richter-4642378 Explore and download more ebooks at ebookbell.com
- 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Selfcontrolled Case Series Studies A Modelling Guide With R Farrington https://ebookbell.com/product/selfcontrolled-case-series-studies-a- modelling-guide-with-r-farrington-7160478 Highenergy Molecular Lasers Selfcontrolled Volumedischarge Lasers And Applications 1st Edition V V Apollonov Auth https://ebookbell.com/product/highenergy-molecular-lasers- selfcontrolled-volumedischarge-lasers-and-applications-1st-edition-v- v-apollonov-auth-5606006 Controlled Drug Delivery The Role Of Selfassembling Multitask Excipients 1st Edition M A Mateescu https://ebookbell.com/product/controlled-drug-delivery-the-role-of- selfassembling-multitask-excipients-1st-edition-m-a-mateescu-4958726 Sequencecontrolled Polymers Synthesis Selfassembly And Properties Jeanfranois Lutz https://ebookbell.com/product/sequencecontrolled-polymers-synthesis- selfassembly-and-properties-jeanfranois-lutz-4920228
- 3. Matlab Model Of An Optimized Battery Charge Controller 1st Edition Subrata Pandey https://ebookbell.com/product/matlab-model-of-an-optimized-battery- charge-controller-1st-edition-subrata-pandey-46413490 Selfoscillations In Dynamic Systems A New Methodology Via Tworelay Controllers 1st Edition Luis T Aguilar https://ebookbell.com/product/selfoscillations-in-dynamic-systems-a- new-methodology-via-tworelay-controllers-1st-edition-luis-t- aguilar-5236250 Controlled Release Technologies 2volume Set Agis F Kydonieus https://ebookbell.com/product/controlled-release-technologies-2volume- set-agis-f-kydonieus-49133498 Controlled Epidemiological Studies Marie Reilly https://ebookbell.com/product/controlled-epidemiological-studies- marie-reilly-49596622 Controlled Process Shooting 1st Edition Joel Turner https://ebookbell.com/product/controlled-process-shooting-1st-edition- joel-turner-50576062
- 7. Urban Maximilian Richter Controlled Self-Organisation Using Learning Classifier Systems
- 9. Controlled Self-Organisation Using Learning Classifier Systems by Urban Maximilian Richter
- 10. KIT Scientific Publishing 2009 Print on Demand ISBN: 978-3-86644-431-7 Diese Veröffentlichung ist im Internet unter folgender Creative Commons-Lizenz publiziert: http://creativecommons.org/licenses/by-nc-nd/3.0/de/ Impressum Karlsruher Institut für Technologie (KIT) KIT Scientific Publishing Straße am Forum 2 D-76131 Karlsruhe www.uvka.de KIT – Universität des Landes Baden-Württemberg und nationales Forschungszentrum in der Helmholtz-Gemeinschaft Dissertation, Universität Karlsruhe (TH) Fakultät für Wirtschaftswissenschaften, 2009 Tag der mündlichen Prüfung: 30. Juli 2009 Referent: Prof. Dr. Hartmut Schmeck Korreferent: Prof. Dr. Karl-Heinz Waldmann
- 13. Controlled Self-Organisation Using Learning Classifier Systems Zur Erlangung des akademischen Grades eines Doktors der Wirtschaftswissenschaften (Dr. rer. pol.) von der Fakultät für Wirtschaftswissenschaften der Universität Karlsruhe (TH) genehmigte DISSERTATION von Dipl.-Wi.-Ing. Urban Maximilian Richter Tag der mündlichen Prüfung: 30. Juli 2009 Referent: Prof. Dr. Hartmut Schmeck Korreferent: Prof. Dr. Karl-Heinz Waldmann 2009 Karlsruhe
- 15. I am not amused about killing so many chickens. IX
- 17. Abstract The complexity of technical systems increases continuously. Breakdowns and fatal errors occur quite often, respectively. Therefore, the mission of organic computing is to tame these challenges in technical systems by providing appropriate degrees of freedom for self-organised behaviour. Technical systems should adapt to changing requirements of their execution environment, in particular with respect to human needs. According to this vision an organic computer system should be aware of its own capabilities, the requirements of the environment, and it should be equipped with a number of so-called self-x-properties. These self-x-properties provide the anticipated adaptiveness and allow reducing the complexity of system management. To name a few characteristics, organic systems should self-organise, self-adapt, self-configure, self-optimise, self-heal, self-protect, or self-explain. To achieve these ambitious goals of designing and controlling complex systems, adequate methods, techniques, and system architectures have to be developed, since no general approach exists to build complex systems. Therefore, a regulatory feedback mechanism is proposed, the so-called generic observer/controller architecture, which constitutes one way to achieve controlled self-organisation in technical systems. To improve the design of organic computing systems, the observer/controller architecture is applied to (generic) multi-agent scenarios from the predator/prey domain. These simple test scenarios serve as testbeds for evaluation. Furthermore, the aspect of (on-line) learning as part of the controller is specially described and the question is investigated, how technical systems can adapt to dynamically changing environments using learning classifier systems as a machine learning technique. Particularly, learning classifier systems are at the focus of many organic computing projects, because they are a family of genetic- and rule-based machine learning methods that fit well into the observer/controller framework. One of their great advantages is that classifier systems aim at the autonomous generation of potentially human-readable results, because they provide a compact generalised representation whilst also maintaining high predictive accuracy. But, learning classifier systems also have drawbacks. The number of reinforcement learning cycles a classifier system XI
- 18. Abstract requires for learning largely depends on the complexity of the learning task. Thus, different approaches to reduce this complexity and to speed up the learning process are investigated and compared. A straightforward way to reduce this complexity is to decompose the task into smaller sub-problems and learn the sub-problems in parallel. It is shown that speeding up the learning process largely depends on the designer’s decision, how to decompose a problem into smaller and modular sub-problems. Thus, different single-agent learning approaches are investigated, which use learning classifier systems that learn in parallel. Furthermore, these parallel learning classifier systems are compared with the organic approach of the two-levelled learning architecture as part of the organic controller. At the on-line level (level 1) the proposed architecture learns about the environment, and about the performance of its control strategies. It does so on-line. Level 2 implements a planning capability based on a simulated model of the environment. At this level the agent can test and compare different alternative strategies off-line, and thus plan its next action without actually acting in the environment. Finally, the potential and relevance of the different learning approaches is evaluated in the case of simple predator/prey test scenarios with respect to more demanding application scenarios. XII
- 19. Acknowledgements Writing scientific publications, especially this thesis, is and has mostly been a lonely business. At the end, only my name will occur on the title. All mistakes and all achievements will be linked to me. However, research does not take place in an evacuated place. There have always been people, supporting my way. Thus, I would like to thank all those people, who have accompanied my way during the last years, months, and weeks and who have given me advice in many kinds so that I have finally accomplished this thesis. Foremost, I would like to thank Prof. Dr. Hartmut Schmeck, my doctoral adviser, for supporting me and my research during the last years. I have greatly benefited from his long experience and his way of leading his research group. He has offered me many degrees of freedom to settle down on research topics that became of my personal interest. I am also extremely thankful for his commitment to timely review this thesis despite his busy timetable. I would also like to thank Prof. Dr. Karl-Heinz Waldmann, from Universität Karls- ruhe (TH), who, without hesitation, accepted the request to serve as second reviewer on the examination committee. Furthermore, many thanks to Prof. Dr. Andreas Oberweis and Prof. Dr. Hagen Lindstädt, both from Universität Karlsruhe (TH), who served as examiner and chairman respectively on the examination committee. I am grateful to my friends and colleagues of the research group Efficient Algo- rithms for the excellent and lively working atmosphere within the team. Special thanks to Jürgen Branke for mentoring my research project and Matthias Bonn for supporting my research by JoSchKa, a really helpful tool to distribute computational intensive simulation tasks among free workstations and servers. Also, many thanks to Andreas Kamper and Holger Prothmann for productive, interesting, and encouraging discussions and reviewing parts of this thesis. Thanks to all others of LS1 for – not only – having funny discussions on lunchtime. Similarly, I am grateful to all external collaborators and project partners. In this context, special thanks to Prof. Dr.-Ing. Christian Müller-Schloer and Moez Mnif, both from Leibniz Universität Hannover. Various parts of this thesis are based on XIII
- 20. Acknowledgements a creative collaboration with them. I have often benefited from common project meetings, their experiences, and their opinion. Moreover, thanks to Emre Çakar, Jörg Hähner, Fabian Rochner, and Sven Tomforde for making travel to Hannover, several workshops, and conferences a lovely and regular experience. Last but not least, I would like to thank my family and friends for their permanent support and being there even in stressful and chaotic times. Special thanks to my parents and my sister, who always believed in me and supported my way in every respect. Thanks to my mother, my sister Helene, and Jan-Dirk for reviewing this thesis concerning language and grammar. Moreover, I am grateful to Niklas, Ana, and mostly Jule for their constant encouragement and for making my Karlsruhe years wonderful and unforgettable. Karlsruhe, October 2009 Urban Maximilian Richter XIV
- 21. Contents List of Tables XIX List of Figures XXI List of Abbreviations XXV 1 Introduction 1 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Objectives and Approach . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.3 Major Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.4 Reader’s Guide to this Thesis . . . . . . . . . . . . . . . . . . . . . . 7 1.5 How this Thesis Was Written . . . . . . . . . . . . . . . . . . . . . . 8 2 Organic Computing (OC) 9 3 Controlled Self-Organisation 13 3.1 Self-Organisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1.1 Understanding Self-Organisation from the Viewpoint of Differ- ent Sciences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1.2 Properties of Self-Organisation . . . . . . . . . . . . . . . . . . 18 3.1.3 Definition of Self-Organisation . . . . . . . . . . . . . . . . . . 20 3.1.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.2 Emergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.3 Architectures for Controlled Self-Organisation . . . . . . . . . . . . . 23 3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4 Observer/Controller Architecture 27 4.1 Observer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.1.1 Model of Observation . . . . . . . . . . . . . . . . . . . . . . . 30 4.1.2 Monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 XV
- 22. Contents 4.1.3 Log File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.1.4 Pre-Processor . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.1.5 Data Analyser . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.1.6 Predictor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.1.7 Aggregator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.1.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.2 Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.2.1 Level 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.2.2 Level 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.3 On-Line Learning and Off-Line Planning Capabilities . . . . . . . . . 46 4.4 Architectural Variants of the Observer/Controller Architecture . . . . 49 4.5 Related Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4.5.1 Autonomic Computing . . . . . . . . . . . . . . . . . . . . . . 52 4.5.2 Operator/Controller Module . . . . . . . . . . . . . . . . . . . 54 4.5.3 Sense, Plan, and Act (SPA) . . . . . . . . . . . . . . . . . . . 57 4.5.4 Component Control, Change Management, and Goal Manage- ment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.5.5 Control Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.5.6 Other Related Approaches . . . . . . . . . . . . . . . . . . . . 65 4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 5 Learning to Control 69 5.1 General Thoughts on Learning . . . . . . . . . . . . . . . . . . . . . . 70 5.2 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 5.3 Learning Classifier Systems (LCSs) . . . . . . . . . . . . . . . . . . . 73 5.3.1 Pittsburgh vs. Michigan Style . . . . . . . . . . . . . . . . . . 74 5.3.2 Single-Step vs. Multi-Step Problems . . . . . . . . . . . . . . 75 5.3.3 Different Implementations . . . . . . . . . . . . . . . . . . . . 77 5.3.4 The eXtended Classifier System (XCS) . . . . . . . . . . . . . 78 5.4 Drawbacks of LCSs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 5.5 Parallelism in LCSs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.5.1 Single-Agent Learning Approach . . . . . . . . . . . . . . . . 84 5.5.2 Multi-Agent Learning Approach . . . . . . . . . . . . . . . . . 88 5.6 Level 2 and Another Covering Method . . . . . . . . . . . . . . . . . 97 5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 6 Test Scenarios 101 6.1 Multi-Agent Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 6.1.1 The Predator/Prey Example . . . . . . . . . . . . . . . . . . . 104 6.1.2 Homogeneous and Non-Communicating Agents . . . . . . . . 104 6.1.3 Heterogeneous and Non-Communicating Agents . . . . . . . . 105 XVI
- 23. Contents 6.1.4 Homogeneous and Communicating Agents . . . . . . . . . . . 105 6.1.5 Heterogeneous and Communicating Agents . . . . . . . . . . . 107 6.1.6 Cooperative and Competitive Multi-Agent Learning . . . . . . 107 6.1.7 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . 109 6.2 Chicken Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 6.2.1 Agent Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . 112 6.2.2 General Simulation Structure . . . . . . . . . . . . . . . . . . 115 6.2.3 Observing the Chickens . . . . . . . . . . . . . . . . . . . . . . 116 6.2.4 Controlling the Chickens . . . . . . . . . . . . . . . . . . . . . 121 6.2.5 Discussion of Special Aspects . . . . . . . . . . . . . . . . . . 128 6.3 Other Multi-Agent Scenarios . . . . . . . . . . . . . . . . . . . . . . . 135 6.3.1 Lift Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . 135 6.3.2 Cleaning Robots . . . . . . . . . . . . . . . . . . . . . . . . . 138 6.3.3 Multi-Rover Scenario . . . . . . . . . . . . . . . . . . . . . . . 139 6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 7 Experimental Design 141 7.1 Design Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 7.2 Pre-Experimental Planning . . . . . . . . . . . . . . . . . . . . . . . . 142 7.2.1 Selection of the Response Variables . . . . . . . . . . . . . . . 142 7.2.2 Choice of Factors, Levels, and Ranges . . . . . . . . . . . . . . 143 7.3 Choice of Experimental Designs . . . . . . . . . . . . . . . . . . . . . 145 8 Results 147 8.1 Preliminary Experiments . . . . . . . . . . . . . . . . . . . . . . . . . 147 8.1.1 Chicken Simulation without Control . . . . . . . . . . . . . . . 148 8.1.2 Parameter Studies Using Single Fixed Rules Controller . . . . 148 8.2 Learning to Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 8.2.1 Effect of Varying the Search Space . . . . . . . . . . . . . . . 157 8.2.2 Effect of Simulation Time . . . . . . . . . . . . . . . . . . . . 159 8.2.3 Effect of Varying Maximal Population Sizes . . . . . . . . . . 160 8.2.4 Effect of Reward Functions . . . . . . . . . . . . . . . . . . . . 161 8.2.5 Effect of Other Parameters as Known from Literature . . . . . 163 8.2.6 Pure On-Line Learning . . . . . . . . . . . . . . . . . . . . . . 164 8.2.7 Learning over Thresholds . . . . . . . . . . . . . . . . . . . . . 166 8.2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 8.3 Parallel XCS Architectures . . . . . . . . . . . . . . . . . . . . . . . . 169 8.3.1 2PXCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 8.3.2 3PXCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 8.3.3 HXCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 8.3.4 Limitations of the Single-Agent Learning Approach . . . . . . 173 8.4 Using Level 2 Learning . . . . . . . . . . . . . . . . . . . . . . . . . . 175 XVII
- 24. Contents 8.5 Using Another Metric on the Observer’s Side . . . . . . . . . . . . . . 177 8.6 Concluding Remarks on the Experiments . . . . . . . . . . . . . . . . 179 9 Conclusion and Outlook 181 9.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 9.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 9.2.1 LCSs as Part of the On-Line Learning Level . . . . . . . . . . 184 9.2.2 Speeding up the Learning Process by Parallelism . . . . . . . 185 9.2.3 Combining On-Line Learning and Off-Line Planning . . . . . . 185 9.2.4 Generality of the Experimental Results . . . . . . . . . . . . . 186 9.3 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 9.3.1 Outlook from the Viewpoint of the Investigated Scenario . . . 187 9.3.2 Outlook from the Viewpoint of the OC Community . . . . . . 188 9.3.3 Outlook from the Viewpoint of the LCSs Community . . . . . 188 9.4 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 References 191 XVIII
- 25. List of Tables 4.1 Comparison of the different levels in the observer/controller architec- ture vs. the operator/controller module . . . . . . . . . . . . . . . . . 57 4.2 Comparison of the different levels in the observer/controller archi- tecture vs. the three-layered architecture from the area of (mobile) robotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.3 Comparison of the different levels in the observer/controller archi- tecture vs. the three-levelled architecture for self-managed software systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 6.1 Parameters of the chicken simulation . . . . . . . . . . . . . . . . . . 114 6.2 Observable parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 118 8.1 Chicken simulation without control . . . . . . . . . . . . . . . . . . . 148 8.2 Combinations of fixed single rules controller parameters . . . . . . . . 150 8.3 Results of the fixed single rule controller experiments over 10 000 ticks with the parameter combination d = 5, ty = 0.2, and th = 0.3 . . . . . 153 8.4 Results of the single fixed rules experiments over 10 000 ticks sorted for the average #kc . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 8.5 Varying values of duration and intensity . . . . . . . . . . . . . . . . 158 8.6 Results of the XCS vs. the best single fixed rules controller established in parameter studies with varying the simulation time . . . . . . . . . 165 8.7 Average number of killed chickens #kc after 10 000 simulated ticks in ascending order using an XCS, which is triggered when predefined thresholds are exceeded . . . . . . . . . . . . . . . . . . . . . . . . . . 167 XIX
- 27. List of Figures 3.1 Simplified view of the generic observer/controller architecture . . . . 14 4.1 Generic observer/controller architecture with two-level learning . . . . 28 4.2 Generic observer architecture consisting of a monitor, a pre-processor, a data analyser, a predictor, and an aggregator . . . . . . . . . . . . . 29 4.3 Example of order perception: Depending on the objective of the observer the nine balls are perceived as orderly or unorderly (position on the left hand side vs. colour on the right hand side) . . . . . . . . 34 4.4 Fingerprint with different attributes at three specific times t0, t1, and t2, visualised as a six-dimensional Kiviat graph (one dimension for each attribute) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.5 Entropy values depending on the probability of the colour red . . . . 38 4.6 Generic controller architecture with two-level learning . . . . . . . . . 43 4.7 Centralised and distributed variants of the generic observer/controller architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4.8 Multi-levelled or hierarchical variant: An observer/controller on each system element as well as one for the whole technical system . . . . . 51 4.9 Structure of an autonomic element, which interacts with other elements and with human programmers via its autonomic manager, see [KC03] 53 4.10 Structure of the operator/controller module, see [HOG04] . . . . . . . 55 4.11 A mobile robot control system is decomposed traditionally into func- tional modules, see [Bro86] . . . . . . . . . . . . . . . . . . . . . . . . 58 4.12 Task achieving behaviours as decomposition criterion for mobile robots, see [Bro86] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 4.13 Control is layered in a hierarchy of levels of competence, where higher layers subsume lower layers in the case of taking control, see [Bro86]: Partitioning the system is possible at every level, the lower layers form a complete operational control system . . . . . . . . . . . . . . . . . . 59 4.14 Three-levelled architecture for self-managed systems, see [KM07] . . . 62 XXI
- 28. List of Figures 5.1 The Woods101 example is a non-Markov environment . . . . . . . . . 77 5.2 Schematic overview of an XCS, see [Wil98] . . . . . . . . . . . . . . . 78 5.3 Variants of parallel LCSs as part of the single-agent learning approach: Parallelism is distinguished on different levels, see [Gia97] . . . . . . . 86 5.4 Population structures for parallel multi-agent LCSs . . . . . . . . . . 89 5.5 Multi-agent learning approach . . . . . . . . . . . . . . . . . . . . . . 91 5.6 Two-level learning architecture is applied to an XCS . . . . . . . . . . 97 6.1 Variants of the predator/prey example, see [SV00] . . . . . . . . . . . 106 6.2 Snapshots of the chicken simulation: Unwounded chickens are white, wounded chickens are dark (red), and feeding troughs are represented by four bigger (yellow) circles. . . . . . . . . . . . . . . . . . . . . . . 110 6.3 An Eurovent cage with 60 chickens . . . . . . . . . . . . . . . . . . . 111 6.4 Finite state machine of a chicken representing the local behaviour rules of a single chicken . . . . . . . . . . . . . . . . . . . . . . . . . . 112 6.5 Operational sequence of the chicken simulation, the contained observ- ing and controlling steps are shown in Figures 6.7 and 6.12 . . . . . . 115 6.6 The generic architecture is applied to the chicken scenario . . . . . . 116 6.7 Steps of the observing process . . . . . . . . . . . . . . . . . . . . . . 117 6.8 Method to predict clustering, see [MMS06] . . . . . . . . . . . . . . . 119 6.9 Emergence value of the x-coordinates over time without any control action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 6.10 Interpolated emergence value of the x-coordinates over time without any control action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 6.11 Number of killed chickens #kc over time (every peak denotes a killed chicken) without control action . . . . . . . . . . . . . . . . . . . . . 121 6.12 Steps of the controlling process . . . . . . . . . . . . . . . . . . . . . 121 6.13 Snapshot of the chicken simulation with noise control . . . . . . . . . 123 6.14 Controlling with fixed single rules . . . . . . . . . . . . . . . . . . . . 124 6.15 Emergence value of the x-coordinates over time with control action . 125 6.16 Interpolated emergence value of the x-coordinates over time with control action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 6.17 Number of killed chickens #kc over time (every peak denotes a killed chicken) with control action . . . . . . . . . . . . . . . . . . . . . . . 126 6.18 If a predefined threshold exceeds, learning will start using an XCS . . 126 6.19 Learning all possible situations using an XCS . . . . . . . . . . . . . 127 6.20 An XCS is equipped with a simulation model on level 2 . . . . . . . . 128 6.21 Simplified chicken scenario . . . . . . . . . . . . . . . . . . . . . . . . 130 6.22 Example with identical entropy and emergence values, respectively . . 131 6.23 An architectural overview of organic traffic control: Level 0 represents the traffic node, levels 1 and 2 are organic control levels responsible for the selection and generation of signal programmes, see [RPB+ 06] . 132 XXII
- 29. List of Figures 6.24 Lifts synchronise, move up and down together, and show the emergent effect of bunching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 7.1 Simplified cause and effect diagram of the chicken simulation . . . . . 144 8.1 Fitness landscape of the chicken simulation depends on three thresholds of critical emergence values and two parameters of a noise signal . . . 149 8.2 Chicken simulation with single fixed rules controller, ty = 0.1, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 151 8.3 Chicken simulation with single fixed rules controller, ty = 0.2, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 151 8.4 Chicken simulation with single fixed rules controller, ty = 0.3, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 151 8.5 Chicken simulation with single fixed rules controller, ty = 0.4, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 152 8.6 Chicken simulation with single fixed rules controller, ty = 0.5, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 152 8.7 Chicken simulation with single fixed rules controller, ty = 0.6, th = 0.3, i ∈ {0, 10, 20, . . . , 50}, and tx ∈ {0.1, 0.2, . . . , 0.6} . . . . . . . . . . . . 152 8.8 Excerpt of a typical XCS’s population . . . . . . . . . . . . . . . . . 155 8.9 Learning condition-action-mappings for all situations . . . . . . . . . 156 8.10 Learning condition-action-mappings for critical situations only . . . . 157 8.11 Learning over time in scenarios with different search spaces, varying parameters of duration and intensity, as shown in Table 8.5, and having a population of maximal 2 500 classifiers . . . . . . . . . . . . 158 8.12 Learning over time in scenarios with different search spaces, varying parameters of duration and intensity, as shown in Table 8.5, and having a population of maximal 5 000 classifiers . . . . . . . . . . . . 159 8.13 Effect of varying the maximal population size . . . . . . . . . . . . . 160 8.14 Effect of varying the reward function . . . . . . . . . . . . . . . . . . 162 8.15 Learning over time using an XCS vs. the best found single fixed rules controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 8.16 XCS over threshold with (tx, ty, th) = (0.1, 0.1, 0.3) vs. XCS . . . . . . 168 8.17 XCS vs. 2PXCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 8.18 Learning over time: XCS vs. 2PXCS, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . . . . . . . 170 8.19 3PXCS vs. HXCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 8.20 Learning over time: XCS vs. 3PXCS, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . . . . . . . 172 8.21 Learning over time: XCS vs. HXCS, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . . . . . . . 173 XXIII
- 30. List of Figures 8.22 Learning over time: XCS vs. XCS with level 2, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . 176 8.23 Comparing learning over time, which is based on different metrics on the observer’s side, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . . . . . . . . . . . . . . . . . 178 8.24 Learning over time: All investigated approaches, averaged values over 20 runs, 15 possible actions, d ∈ {5, . . . , 9} × i ∈ {0, 10, 20} . . . . . . 179 XXIV
- 31. List of Abbreviations #kc – The number of killed chickens 2PXCS – Two parallel instances of the extended classifier system 3PXCS – Three parallel instances of the extended classifier system d, dj – The duration of a noise signal ex, ey, eh – The relative emergence indicators HXCS – The hierarchical organised extended classifier system i, ij – The intensity of a noise signal LCS – A learning classifier system L2 – Learning (or planning) on level 2 of the generic observer/controller archi- tecture OC – Organic computing SuOC – The system under observation and control tx, ty, th – Predefined thresholds of critical emergence values XCS – The extended classifier system, as introduced in [Wil95] XXV
- 33. Chapter1 Introduction In the Nevada desert, an experiment has gone horribly wrong. A cloud of nanoparticles – micro-robots – has escaped from the laboratory. This cloud is self-sustaining and self-reproducing. It is intelligent and learns from experience. For all practical purposes, it is alive. It has been programmed as a predator. It is evolving swiftly, becoming more deadly with each passing hour. Every attempt to destroy it has failed. And we are the prey. As fresh as today’s headlines, Michael Crichton’s most compelling novel yet tells the story of a mechanical plague and the desperate efforts of a handful of scientists to stop it. Drawing on up-to-the-minute scientific fact, Prey takes us into the emerging realms of nanotechnology and artificial distributed intelligence – in a story of breathtaking suspense. Prey is a novel you can’t put down. Because time is running out. [Cri02] These words cited above are written on the hardcover version of the techno-thriller novel Prey by Michael Crichton. Of course, the book is science fiction and human beings fighting against a swarm of micro-robots seems to be not realistic so far, but an interesting story is told that features relatively new advances in computer science, such as artificial life, swarm intelligence, self-organisation, genetic algorithms, or multi-agent-based computing. Major themes of the book deal with the threat of intelligent micro-robots escaping from human control and becoming autonomous, self-replicating, and, by that, dangerous. Many aspects of the story, such as the cloud-like nature of the nanoparticles and their nature-inspired process of evolution closely follow research done by computer scientists in the past (few) years, see e. g., the fields of evolutionary computation [FOW66, Gol89, Hol75], computational intelligence [Eng02, PMG98], or artificial life1 . 1 http://www.alife.org
- 34. Chapter 1 Introduction As short-sighted decision-making at the corporate level can lead to a disaster when the companies involved control dangerous new technology, the book is about the potential consequences, if suitable controls are not placed on biotechnology, before it will develop to such an extent that it can threaten the survival of life on earth. Of course, this is an important discussion and scientists doing research in informatics or biologically inspired informatics have to cope with it. Hence, this thesis will focus on the challenge of designing technical systems, which are inspired by swarm intelligence, multi-agent systems, or self-organisation and enable controllability of these systems at the same time. The research on these different domains has intensified. A growing number of conferences2 , workshops3 , and journals4 supports this trend. Paradigms of computing are emerging based on modelling and developing compu- ter-based systems exploiting ideas that are observed in nature. The human body’s autonomic nervous system inspires the design of self-organising computer systems, as proposed in IBM’s autonomic computing initiative5 . Some evolutionary systems are modelled in analogy to colonies of ants or other insects. Highly-efficient and highly-complex distributed systems are developed to perform certain functions or tasks using the behaviour of insects as inspiration, e. g., swarms of bees, flocks of birds, schools of fish, or herds of animals. Self-organising systems are not science fiction any more, but problems with in- creasing complexity and controllability of technical systems call for new system architectures, as postulated in the field of organic computing (OC) and explicitly investigated in this thesis. 1.1 Motivation As mentioned in [BMMS+ 06], the impressive progress in computing technology over the past decades has not only led to an exponential increase in available computing power, but also to a shrinking of computer chips to a miniature format. While only twenty years ago, the predominant computing platform was a company mainframe shared by many users, today, a multitude of embedded computing devices surrounds us, including PDA, cell phone, digital camera, navigation system, MP3-player, etc. in everyday life. An additional trend during recent years has been that these devices are equipped with (often wireless) communication interfaces, allowing them to interact and exchange information. 2 E. g., the International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS) or the Genetic and Evolutionary Computation Conference (GECCO) 3 E. g., the International Workshop on Learning and Adaptation in Multi-agent Systems (LAMAS) or the International Workshop on Learning Classifier Systems (IWLCS) 4 E. g., Artificial Life (MIT Press Journals) or ACM Transactions on Autonomous and Adaptive Systems (TAAS) 5 http://www.research.ibm.com/autonomic or see Section 4.5.1 2
- 35. 1.1 Motivation Amongst others this outlook towards smaller, more intelligent, and more numer- ous devices surrounding everybody in his everyday life is given by the paradigm of ubiquitous computing, which was first introduced by Mark Weiser in [Wei91]. Future information processing will be integrated into a broad range of everyday and everywhere objects making these objects intelligent. These devices will be interconnected and they will communicate over various communication channels. Thus, networks of intelligent systems will grow, and their behaviour will no longer be predictable with certainty due to interaction effects, see [Sch05a]. In addition, other large technical systems consist of more and more interconnected electronic devices. For example, in cars, numerous processors and embedded systems keep the vehicle on the road, control the engine with respect to combustion and pollution, assist the driver, provide security with air bags and seat belt systems, provide functions such as air conditioning, navigation, parking assistant, information services, and entertain the passengers. All these controllers are connected to a complex communication network. And this development has not stopped yet. Technical innovations are only a stone’s throw away from scenarios like smart factory, with flexible robots self-organising to satisfy the needs at hand [Gui08], or smart cars that adapt to different drivers and road conditions, communicate with other cars on special events, or integrate personal devices (PDA, mobile telephone, or notebook) into their network. While this development is exciting, the resulting systems become increasingly complex, up to the point where they can no longer be designed or used easily. Even today, in the automotive sector, it is estimated that about half of all car break-downs are caused by electric and electronic components. E. g., in 2005 weak car batteries head the table of causes for break-downs listed by the ACE Auto Club Europa with 25% [ACE06], while electronic components are listed on third position (currently) with 13%, still increasing their percentage. Thus, the questions arise, how to design such complex distributed and highly interconnected systems, and how to make them reliable and usable. Clearly, the designer is not able to foresee all possible system configurations, and to prescribe proper behaviours for all cases. Additionally, the user is relieved from having control in detail over all parameters of the system, allowing him or her to influence the system on a higher level, e. g., by setting goals. OC has the vision to meet this challenge of coping with increasing complexity by making technical systems more life-like, and endowing them with properties such as self-organisation, self-configuration, self-repair, or adaptation [Sch05b]. Future systems will possess certain degrees of freedom to handle unforeseen situations and act in a robust, flexible, and independent way. Thus, these systems will exhibit self-organised behaviour, which makes them able to adapt to a changing surrounding. That is why, in the field of OC, these systems are called organic. Hence, an OC system is a system, which dynamically adapts to the current situation of its environment, but still obeys goals set by humans. In addition to this environmental awareness, systems 3
- 36. Chapter 1 Introduction providing services for humans will adjust themselves to the users’ requirements (and not vice versa). Based on these trends, the question, also addressed in this thesis, is not, whether complexity increases or informatics is confronted with emergent behaviour, but how new technical systems will be designed that have the possibility to cope with the emerging global behaviour of self-organising systems by adequate control actions. 1.2 Objectives and Approach As outlined before and motivated in Chapter 2, OC has a major research interest in new system architectures that self-organise and adapt exploiting certain degrees of freedom. To achieve these ambitious goals of designing and controlling self-organising systems, adequate methods and system architectures have to be developed, since no general approach exists to build OC systems. Therefore, OC proposes a regulatory feedback mechanism, the so-called generic observer/controller architecture [MS04], which constitutes one way to achieve controlled self-organisation in technical systems. Using this control loop, an organic system will adapt over time to its changing environment. It is obvious that this architecture could benefit from learning capabil- ities to tackle these challenges. Therefore and as described in detail in Chapter 4, the controller has been refined by a two-levelled learning approach. At the on-line level (level 1) the proposed architecture learns about the environment, and about the performance of its control strategies. Level 2 implements a planning capability based on a simulated model of the environment. At this level an agent can test and compare different alternative strategies off-line, and plan its next action without actually acting in the environment. Thus, the two research questions addressed in this thesis are defined in the following in more detail. 1. What does it mean to establish and utilise controlled self-organisation in the context of technical OC scenarios specially focussing on learning classifier systems (LCSs) as machine learning technique on level 1 of the proposed two-levelled learning architecture? 2. How is the (on-line) learning process speeded up? In other words, the observer/controller architecture is refined into a form that can serve as a generic template containing a range of components, which should be necessary in a range of OC application scenarios. It enables a regulatory feedback mechanism and the use of machine learning techniques to improve a single-agent’s or a multi-agent system’s behaviour in technical domains with the following characteristics. • There exists a need for on-line decision-making, • decisions are based on (aggregated) sensor information, 4
- 37. 1.2 Objectives and Approach • decisions are influenced by and based on decisions that have been taken by other agents, and • several agents act with a cooperative/competitive, well-defined, and high-level goal. Thus, the agents are assumed to have the following characteristics. • The ability to process, aggregate, and quantify sensor information, • the ability to use this information to update and control their (local) behaviour, and • the ability to cope with limited communication capabilities, e. g., caused by local neighbourhoods, low bandwidth, power restrictions, etc. In the following, scenarios are mainly investigated, where a collection of (non- adaptive) agents is observed and controlled by a centralised observer/controller architecture. In these scenarios, learning takes place on a higher level of abstraction. The general approach to answering the thesis questions has been to investigate selected ideas of the generic observer/controller architecture within different multi- agent scenarios, which serve as representative OC test scenarios. Since the main goal of any testbed is to facilitate the trial and evaluation of ideas that show great promise for real world applications, e. g., smart production cells, smart factories, logistics, traffic, automotive industry, or information technology, the chosen test scenarios are assumed to have the following properties. • To allow for generalisation of the results, each test scenario should exhibit a different emergent phenomenon, which could be observed and controlled, hence justifying the utility of the observer/controller architecture. • On the other hand, the test scenarios should be rather simple to implement and easy to understand. Therefore, all of the thesis contributions have originally been developed in simulated scenarios of the predator/prey domain, which has served as demonstrating and evaluating scenario for manifold research ideas for a long time. An initial assumption was that in domains with the above characteristics agents should map their sensor information to control actions. LCSs could provide such a mapping. Therefore, their suitability had to be investigated. The use of LCSs is specially focussed on level 1 of the proposed two-levelled learning architecture and methods are successfully contributed, which equip such multi-agent scenarios, as mentioned above, with OC ideas. 5
- 38. Chapter 1 Introduction While LCSs have drawbacks in learning speed, which seem to be critical in combination with technical applications, mechanisms have been investigated, which speed up the learning process of LCSs. These approaches are compared with the proposed two-levelled learning architecture that learns on-line (level 1) about the environment and about the performance of its control strategies, while on level 2 a planning capability is used, based on a simulation model of the environment, where an agent can test and compare different alternative strategies off-line, and thus plan its next action without actually acting in the environment. 1.3 Major Contributions In brief, this thesis makes three main contributions related to the research fields of OC. First, OC research is summarised that has been done over the last five years and specially focusses on the design of the generic observer/controller architecture, which serves as a framework for building OC systems. This architecture allows for self-organisation, but at the same time enables adequate reactions to control the – sometimes completely unexpected – emerging global behaviour of these self- organised technical systems. The proposed architecture can be used in a centralised, distributed, or multi-levelled and hierarchically structured way to achieve controlled self-organisation as a new design paradigm. Thus, Chapters 3 and 4 address related work in the field of architectures for controlled self-organisation. Some contents have been published in [ÇMMS+ 07, SMSÇ+ 07, SMS08]. Chapter 4 mainly bases on [BMMS+ 06, RMB+ 06]. Secondly, the idea of a two-level learning approach is introduced as part of the controller. Since a learning capability is an essential feature of OC systems, the generic architecture and, in particular, the controller, has to include adequate components for learning. The work, presented in this thesis, focusses on on-line learning and specially on the investigation of LCSs as an adequate machine learning technique. Thirdly, several distributed variants of LCSs are investigated with the objective of improving learning speed and effectiveness. While conventional LCSs have drawbacks in learning speed, this thesis investigates possible modifications by decomposing a problem into smaller sub-problems and by learning these subtasks independently. Furthermore, the performance of these variants of on-line LCSs are compared to the combination of on-line learning and off-line planning capabilities, as suggested by the observer/controller architecture. Since the second and the third contributions are inherently domain-specific, the following chapters provide a general specification as well as an implementation within multi-agent test scenarios. The used multi-agent test scenarios are selected from the predator/prey domain and can therefore be generalised to other domains. Also, work, which has been done in [RRS08], is shortly summarised in Chapter 6. Chapters 5, 6, and 8 present results that have been published in [MRB+ 07, RM08, RPS08]. 6
- 39. 1.4 Reader’s Guide to this Thesis In Chapter 9, an extended review of the empirical results validating the major contributions of this theses is given. 1.4 Reader’s Guide to this Thesis To enjoy oneself reading this thesis and to identify the most relevant chapters from a personal point of view, a general description of the contents of each chapter should guide the reader. The presented work is structured as follows. • Chapter 2 summarises the vision of OC, since this thesis is mainly based on OC research topics and copes with an interdisciplinary view, unknown in literature before, which connects different research fields, e. g., control theory, machine learning, or multi-agent theory. • In Chapter 3 related work concerning the topic of controlled self-organisation is reviewed. Self-organisation and emergent phenomena have been research topics in several areas. The most relevant ideas are summed up with regard to this thesis. • In Chapter 4 the generic observer/controller architecture is introduced, which serves as a framework to build OC systems. The presented work is focussed on the centralised variant of this design paradigm and every module of this architecture is explained in detail. To compare the organic approach to other regulatory feedback mechanisms, the two-level learning approach as part of the controller is specially described. • Since the capability to adapt to dynamically changing environments is in the main focus of OC systems, the aspect of learning is investigated in detail. Thus, LCSs are presented in Chapter 5. This chapter reviews the state of the art from LCS’s literature and defines the idea of parallel classifier systems to speed up the learning process. • Chapter 6 introduces the domains used as test scenarios within the thesis. A nature-inspired scenario has been implemented and serves as a testbed to validate the learning cycle of a centralised observer/controller architecture. • General design decisions concerning the implemented learning architectures with respect to the nature-inspired test scenario are outlined in Chapter 7. The actual analysis of the results is given in Chapter 8. • Chapter 9 summarises the contributions of this thesis and outlines the most promising directions for future work. Since several chapters of this thesis contain their own related work sections describing the research most relevant 7
- 40. Chapter 1 Introduction to their contents, this chapter is used for a survey about OC from an LCS’s perspective. 1.5 How this Thesis Was Written This thesis is the outcome of several years of research with financial support by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) within the priority programme 1183 OC. Several papers have been published with different colleagues in a close cooperation between the research group of my doctoral adviser, Prof. Dr. Hartmut Schmeck, and the group of Prof. Dr.-Ing. Christian Müller-Schloer from Leibniz Universität Hannover and were taken as the basis for the following chapters. For reasons of presentation, the chronological order, in which the articles appeared, does not coincide with the presented order within the following chapters. 8
- 41. Chapter 2 Organic Computing (OC) It is not the question, whether adaptive and self-organising systems will emerge, but how they will be designed and controlled. [Sch05b, SMSÇ+ 07] Since the work presented here is mainly based on OC research topics, this chapter summarises the vision of OC. Motivation and challenges of this young research field are explained in short, before the contributions of this thesis are described in the following chapters. As outlined in Section 1.1, the increasing complexity of technical systems calls for research into new design principles. It is impossible for a designer to foresee all possible configurations and to explicitly specify the entire behaviour of a complex system on a detailed level. In particular, if the system consists of many interacting components, it may exhibit new, emergent properties that are very difficult to anticipate. Emergent phenomena are often identified, when the global behaviour of a system appears more coherent and directed than the behaviour of individual parts of the system (the whole is more than the sum of its parts). More generally, such phenomena arise in the study of complex systems, where many parts interact with each other and where the study of the behaviour of individual parts reveals little about system-wide behaviour. Especially, in the area of multi-agent systems emergence and self-organisation have been studied extensively, see [DFH+ 04, DGK06] for two recent surveys. Despite their complexity, living creatures are very robust and have the natural ability to learn and adapt to an uncertain and dynamic environment. The idea of OC is therefore to address complexity by making technical systems more life-like and to develop an alternative to the explicit total a priori specification of a system. Instead, organic systems should adapt and self-organise with respect to some degrees of freedom. But, OC systems should be designed with respect to human needs, and have to be trustworthy, robust, adaptive, and flexible. They will show the so-called self-x-properties: Self-configuration, self-optimisation, self-healing, self-explanation,
- 42. Chapter 2 Organic Computing (OC) and self-protection. Such systems are expected to learn about their environment during life time, will survive attacks and other unexpected breakdowns, will adapt to their users, and will react sensibly, even if they encounter a new situation, for which they have not been programmed explicitly. In other words, an OC system should behave more life-like (organic). This can only be achieved by adding some kind of awareness of their current situation to the system elements and the ability to provide appropriate responses to dynamically changing environmental conditions. The principles of OC are strongly related to the objectives of IBM’s autonomic computing initiative, see Section 4.5.1. But, while autonomic computing is directed towards maintaining server architectures, which should be managed without active interaction between man and machine [KC03, Ste05], OC’s focus is more general in its approach and addresses large collections of intelligent devices, providing services to humans, adapted to the requirements of their execution environment [Sch05b]. Thus, besides showing the self-x-properties, interaction between man and machine is an essential part of OC systems. The term organic computing was formed in 2002 as a result of a workshop aiming at future technologies in the field of computer engineering. The outlines of the workshop and the OC vision were first formulated in the joint position paper [ACE+ 03] of the section of computer engineering (Technische Informatik) of the Gesellschaft für Informatik (German Association for Informatics, GI) and the Informationstechnische Gesellschaft (German Association for Information Technology). In 2005, the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) approved a priority research programme on OC for six years (2005–2011). This research programme addresses fundamental challenges in the design of OC systems; its goal is a deeper understanding of emergent global behaviour in self-organising systems and the design of specific concepts and tools to construct and control OC systems for technical applications. Topics, such as adaptivity, reconfigurability, emergence of new properties, and self-organisation, play a major role. Currently, the research programme provides funding for 18 research projects with a total volume of around EUR 2 million per year. The topics of these projects range from traffic control over robot coordination to chip design. Information on the different projects can be found via the OC website1 . Self-organising systems bear several advantages compared to classical, centrally controlled systems. Amongst others, the failure of a single component should not cause a global malfunction of the whole system. Such a system will be able to adapt to changing circumstances. As a result, self-organisation could be described as a method of reducing the complexity of computer systems. In such self-organising systems the local interaction of the system elements may result in an emergent global behaviour, which can have positive (desired) as well as 1 http://www.organic-computing.de/spp 10
- 43. negative (undesired) effects. Self-organisation and emergent phenomena also initiate new problems unknown in the engineering of classical technical systems. A global emergent behaviour usually is a nonlinear combination of local behaviours. Its design process with both potential design directions (top-down vs. bottom-up design) turns out to be a highly non-trivial task: For a top-down approach it is hard to deduce adequate local rules from a desired global behaviour, and in the bottom-up direction it quite often remains unclear how local rules go together with global behaviour, see [KC03]. In this context and in order to assess the behaviour of the technical system and – if necessary – for a regulatory feedback to control its dynamics, the so-called observer/controller architecture has become widespread in the OC community as a design paradigm to assure the fulfilment of system goals (given by the developer or user), see Chapter 4. The observer/controller uses a set of sensors and actuators to measure system variables and to influence the system. Together with the system under observation and control (SuOC), the observer/controller forms the so-called organic system. An observer/controller loop enables adequate reactions to control the – sometimes completely unexpected – undesired emerging global behaviour resulting from local agents’ behaviour. However, besides this fascinating outlook, the materialisation of the OC vision depends on several crucial factors, which are summarised in [Sch05b]. • Designers of OC systems have to guarantee that self-organising systems, based on OC principles, do not show unwanted (emergent) behaviour. This is particularly important, when malfunction can have disastrous consequences, e. g., in safety critical applications. The generic observer/controller architecture, as described in Chapter 4, seems to be a promising approach in asserting certain functionality and additionally in keeping the system at an effective state of operation. OC systems will only become accepted, if users can trust them. Therefore, trust/reliability could turn out to be the most important prerequisite for acceptance. • Closely related is the need for the user to monitor and influence the system: It has to be guaranteed that it is still the user, who guides the overall system. Therefore, the system developer has to design user interfaces, which can be used to control the system – which means that there has to be a possibility to take corrective actions from outside the system. The generic observer/controller architecture considers this requirement. • Developers of OC systems have to determine appropriate rules and patterns for local behaviour in large networks of smart devices in order to provide some requested higher functionality. Important topics in the design of self-organising systems are the utilisation of arising emergent phenomena and controlling the local level in such a way that the system shows the desired behaviour at global 11
- 44. Chapter 2 Organic Computing (OC) level. Therefore, the task is to derive a set of behavioural and interaction rules that, if embedded in individual autonomous elements, will induce a certain global characteristic. The inverse direction of anticipating the global system behaviour based on known local decision rules is also very important in this regard. • On interacting with humans an OC system has to show context sensitive characteristics and has to filter information and services according to the current situation or user’s needs. In general, an OC system has to be aware of its environment and should act accordingly. • Building OC systems, one has to think carefully about how to design the necessary degrees of freedom for the intended adaptive behaviour. Certain degrees of freedom are needed to enable self-organisation, but it is easily imaginable that allowing the different parts of a system a broad range of possible (re-)actions in a specific situation could result in uncontrollable chaos. • The implementation of learning abilities as part of OC systems provides great chances, but also bears several problems. Learning systems often learn from mistakes. In fact, they will make mistakes, if no countermeasures are taken. Additionally, developers can guide the learning process of a learning system and they can assure that the system does not develop itself in an unwanted (emergent) manner. This aspect of learning is focussed on in detail in Chapter 5. This list could be considerably expanded, but it already represents the most important topics. The following chapter will concentrate on the organic vision of controlled self-organisation. 12
- 45. Chapter 3 Controlled Self-Organisation Technological systems become organised by commands from outside, as when human intentions lead to the building of structures or machines. But, many natural systems become structured by their own internal processes: These are self-organising systems, and the emergence of order within them is a complex phenomenon that intrigues scientists from all disciplines. [YGWY88] Self-organising systems are well known from nature and have been studied in domains like physics, chemistry, and biology. In recent years research interest has succeeded to apply concepts of self-organisation to technical systems. The reason is a paradigm shift from monolithic systems to large networked systems driven by the technological change of integrating more information processing into the everyday life, objects, and activities. The necessity to find new approaches to cope with upcoming problems of increasing complexity attracts awareness to the principle of self-organisation. OC systems should use self-organisation to achieve a certain externally provided goal. Furthermore, the system has to adapt to changing environmental requirements and to be capable to deal with (unanticipated) undesired emergent behaviour. Therefore, OC systems are assumed to support controlled self-organisation. Whenever necessary, this requires a range of methods for monitoring and analysing the system performance and for providing appropriate control actions. The generic observer/controller architecture – this architecture is introduced in detail in Chapter 4 – promises to provide the necessary components for satisfying all these demands, see Figure 3.1. Similar to the MAPE cycle (monitor, analyse, plan, and execute) of IBM’s au- tonomic computing initiative, see Section 4.5.1, a closed control loop is defined to keep the properties of the self-organising SuOC within preferred boundaries. The observer observes certain (raw) attributes of the system and aggregates them to situation parameters, which concisely characterise the observed situation from a
- 46. Chapter 3 Controlled Self-Organisation organic system SuOC goals : agent/robot/entity system status observer controller observes controls reports input output learning Figure 3.1: Simplified view of the generic observer/controller architecture global point of view, and passes them to the controller. The controller acts according to an evaluation of the observation (which might include the prediction of future behaviour). If the current situation does not satisfy the requirements, it will take action(s) to direct the system back into its desired range, will observe the effect of the intervention(s), and will take further actions, if necessary. Using this control loop an organic system will adapt over time to its changing environment. It is obvious that the controller could benefit from learning capabilities to tackle these challenges. Although the observing and controlling process is executed in a continuous loop, and the SuOC is assumed to run autonomously, even if the observer/controller archi- tecture is not present – even though in a suboptimal way. Furthermore, emergence plays a central role in OC systems. Emergent and self-organising behaviour has been observed in nature, demonstrated in a variety of computer simulated systems in artificial life research, and it has also occurred in highly complex technical systems, where it has quite often led to unexpected global functionality [BDT99]. Despite the importance of a rigorous description of these phenomena, the quantitative analysis of technical self-organising systems is still a rather unexplored area. Therefore, this chapter describes the understanding of the basic mechanisms of self-organisation and emergent behaviour in complex (organic) ensembles, summarises related work, and provides appropriate (metrics and) tools for utilising controlled self-organisation. In Section 3.1 research is summarised that has been done in the area of self-or- ganisation and in Section 3.2 about the related concept of emergence. There may be instances of self-organisation without emergence and emergence without self- organisation, and there is evidence in literature that the phenomena are not the same. 14
- 47. 3.1 Self-Organisation However, future research is needed to clarify the relation between these two terms. Finally, in Section 3.3 an architectural-based approach for the design and engineering of technical systems is proposed that makes use of controlled self-organisation, before describing the OC approach in Chapter 4. 3.1 Self-Organisation The dynamics of a system can tend by themselves to increase the inherent order of a system. This idea has a long history, being first introduced by the French philosopher René Descartes. In 1947, the term self-organisation was introduced by the psychiatrist and engineer William Ross Ashby [Ash47]. Cyberneticans, like Heinz von Foerster, Stafford Beer, Gordon Pask, and Norbert Wiener, took up this concept and associated it with general systems theory in the 1960ies. In the 1970ies and 1980ies physicists adopted self-organisation to the field of complex systems and established the topic in scientific literature. Even if the concept of self-organisation is very promising to solve complex problems, as explained in [Ger07], the notion of self-organisation will remain somewhat vague, and discussion has been widespread. The extensive FAQ-list1 is a good link to research that has been done so far. The term self-organisation is used frequently, but a generally accepted meaning has not emerged. As the list grows, it becomes increasingly difficult to decide whether these phenomena are all based on the same process, or whether the same label has been applied to several different processes. Despite its intuitive simplicity as a concept, self-organisation has proven notoriously difficult to describe and define formally or mathematically. Thus, it is entirely possible that any precise definition might not include all the phenomena, to which the term of self-organisation has been applied. In the following, it will not be attempted to give a new definition facing the philosophical problem of defining self, the cybernetic problem of defining system, or the universal problem of defining organisation. Instead, research is summarised that has been done so far to characterise the conditions necessary to call a system self-organising. Answers will be given to the following questions: What is a self- organising system? What is it not? And what are possible approaches to engineer self-organising technical systems? 3.1.1 Understanding Self-Organisation from the Viewpoint of Different Sciences As pointed out by Carlos Gershenson in [Ger07], the term self-organisation has been used with different meanings, e. g., in computer science [HG03, MMTZ06], biology [CDF+ 03, FCG06], mathematics [Len64], cybernetics [Ash62, von60], synergetics 1 http://www.calresco.org/sos/sosfaq.htm 15
- 48. Chapter 3 Controlled Self-Organisation [Hak81], thermodynamics [NP77], complexity [Sch03], information theory [Sha01], and evolution of language [de 99]. Selected ideas, which specially contribute to the idea of OC, are summarised from the viewpoint of different sciences. Self-Organisation in Nature According to [CDF+ 03], self-organisation in biological systems is often described as a process, in which a pattern at the global level of a system solely emerges from numerous interactions among the lower level components of the system. Moreover, the rules specifying interactions among the system’s components are executed using only local information, without reference to the global pattern [YGWY88]. A self-organising system in nature acts without centralised control and operates according to local contextual information. Thus, spontaneous behaviour without external control produces a new organisation reacting to environmental changes/dis- turbances. Natural systems often show a robust behaviour, they adapt to changes, and they are able to ensure their own survivability. There are quite a few examples of natural systems, which are not at all robust (in particular at the individual level) and which cannot adapt to changes. Robustness often is a property of a population/swarm, not of an individual. In some cases, self-organisation is linked to emergent behaviour, as described later in Section 3.2. Individual components carry out a simple task, and as a whole these components are able to carry out a complex task emerging in a coherent way through the local interactions of various components. Typical examples from nature are found in the following. • Social insects, like ants, termites, or honey bees, where communication oc- curs through stigmergy by placing chemical substances, pheromones, into the environment. In 1959 the theory of stigmergy has been defined as the work excites the workers by [Gra59]. Direct interactions between the individuals are not necessary to coordinate a group. Indirect communications between the individuals and the environment are enough to create structures. Thus, coordination or regulation tasks are achieved without centralised control. • Flocks of birds or schools of fish, where collective behaviour is defined by simple rules, like getting close to a similar bird (or fish) – but not too much – and getting away from dissimilar birds (or fishes) to collectively avoid predators. • Social behaviours of humans, where emergent complex global societies arise by working with local information and local direct or indirect interactions. • The immune system of mammalians, where cells regenerate self-organised. 16
- 49. 3.1 Self-Organisation From a very general point of view, the notion of autopoiesis is often associated with the term of self-organisation. In the 1970ies biological studies established autopoiesis (meaning self-production) [MV91, Var79], which describes the process of a living system as an organisation to produce itself. An autopoietic system is autonomous and operationally closed, in the sense that every process within it directly helps maintaining the whole. For example, cells or organisms self-maintain the system through generating system’s components. The cell is made of various biochemical components such as nucleic acids and proteins, and is organised into bounded structures such as the cell nucleus, various organelles, a cell membrane, and cytoskeleton. These structures, based on an external flow of molecules and energy, produce the components, which, in turn, continue to maintain the organised bounded structure that gives rise to these components. In computing, an analogous concept is called bootstrapping, which refers to tech- niques that allow a simple system to activate a more complicated system, e. g., when starting a computer a small programme, the built in operating system (BIOS), initialises and tests the hardware (peripherals or memory), loads another programme, and passes control to this programme (an operating system). Self-Organisation in Chemistry Self-organisation is also relevant in chemistry, where it has often been taken as being synonymous with self-assembly. To name a few, this includes research in molecular self-assembly, which is the process, by which molecules adopt a defined arrangement without guidance or management from an outside source [Leh88, Leh90]. Additionally, self-organisation is used in the context of reaction diffusion systems, which are mathematical models that describe, how the concentration of one or more substances is distributed in space changes. This occurs under the influence of two processes that are local chemical reactions, in which the substances are converted into each other, and diffusion, which causes the substances to spread out in space [Fif79]. Other examples are autocatalytic networks, or liquid crystals. Through thermodynamics studies [GP71] the term self-organisation itself has been established in the domain of chemistry in the 1970ies. When an external energy source is applied to an open system, this system decreases its entropy (where order comes out of disorder, see Section 4.1.5). In other words, a system reaches a new system state, where entropy is decreased, when external pressure is added. Compared to the stigmergy concept, mentioned in Section 3.1.1, where self-organisation results from a behaviour occurring inside the system (from the social insects themselves placing pheromones in the environment), this is a fundamental difference from the understanding in chemistry. In the latter case, self-organisation seems to be a result of an external pressure applied from the outside. 17
- 50. Chapter 3 Controlled Self-Organisation Self-Organisation in Mathematics and Computer Science Self-organisation has also been observed in mathematical systems such as cellular automata [TGD04]. In computer science, some instances of evolutionary computation and artificial life exhibit features of self-organisation. Research in artificial systems has been oriented towards introducing self-organi- sation mechanisms specifically for software applications, see [BDHZ06, BDKN05, BHJY07, DKRZ04]. These applications have been inspired by already mentioned nature-inspired concepts like stigmergy, autopoiesis, or the holon concept introduced by Artúr Kösztler in [Kös90]. The term holon describes systems, which represent whole systems and parts of larger systems at the same time. Then, holarchies describe hierarchies of such holons. Typical examples of self-organising artificial systems are swarm-inspired techniques for routing [BDT99] or load-balancing [MMB02]. Furthermore in multi-agent systems, (software) agents play the role of self- organising autonomous entities. Frequently, multi-agent systems are used for simulat- ing self-organising systems, in order to get a better understanding of the dependencies in such systems or to establish models of the simulated systems. As mentioned in [DGK06], the tendency of initiatives like OC or autonomic computing is now to shift the role of agents from simulation to the development of distributed systems. Components (e. g., software agents) that once deployed self-organise in a predefined environment and work in a distributed manner towards the realisation of a given (global) possibly emergent functionality. 3.1.2 Properties of Self-Organisation In general, the understanding of self-organisation seems to be widespread. According to [DGK06] self-organisation essentially refers to a spontaneous and dynamically produced (re-)organisation. Several, more qualitative, properties/issues should be extracted from the different viewpoints mentioned above in the following section to end up in a possible definition in Section 3.1.3. Properties of Self-Organisation in Nature According to swarm intelligence [BDT99], self-organising processes are characterised by four properties. 1. Multiple interactions among the individuals, 2. retroactive positive feedback (e. g., increase of pheromone, when food is de- tected), 3. retroactive negative feedback (e. g., pheromone evaporation), and 18
- 51. 3.1 Self-Organisation 4. increase of behaviour modification (e. g., increase of pheromone, when new path is found). From the more biologically-inspired viewpoint of autopoiesis, a self-organising system could be characterised as an autopoietic machine, which is a machine that is organised (defined as a unity) as a network of processes of production (transformation and destruction) of components, which 1. through their interactions and transformations continuously regen- erate and realise the network of processes (relations) that produced them; and 2. constitute it (the machine) as a concrete unity in space, in which they (the components) exist by specifying the topological domain of its realisation as such a network [MV91]. Properties of Self-Organisation in Chemistry Under external pressure, self-organising behaviour is characterised by a decrease of entropy and satisfies the following requirements, as stated in [GP71]. 1. Mutual causality: At least two components of the system have a circular relationship, each influencing the other. 2. Autocatalysis: At least one of the components is causally influenced by another component, resulting in its own increase. 3. Far from equilibrium condition: The system imports a large amount of energy from outside the system, uses the energy to help renew its own structures (autopoiesis), and dissipates rather than accumulates, the accruing disorder (entropy) back into the environment. 4. Morphogenetic changes: At least one of the components of the system [has to] be open to external random variations from outside the sys- tem. A system exhibits morphogenetic change when the components of the system are changed themselves. Properties of Self-Organisation in Artificial Systems In [DGK04], two definitions of self-organisation in artificial systems have been established. Self-organisation implies organisation, which in turn implies some ordered structure as a result of component behaviour. A new distinct organisation is self-produced, since the process of self-organisation changes the respective structure and behaviour of a system. 19
- 52. Chapter 3 Controlled Self-Organisation • Strongly self-organising systems are systems that change their or- ganisation without any explicit, internal or external, central control. • Weakly self-organising systems are systems where reorganisation occurs as a result of internal central control or planning. 3.1.3 Definition of Self-Organisation The previous sections have shown that it is not trivial to give a precise definition of self-organisation. However, a practical notion, as given in [Ger07], will suffice for the purposes of this thesis. A system described as self-organising is one, in which elements interact in order to dynamically achieve a global function or behaviour. Furthermore, this self-organising behaviour is autonomously achieved through distributed interactions between the system components, which produce feedbacks that regulate the system. Instead of answering the question, which are the necessary conditions for a self-organising system, another question can be formulated: When is it useful to describe a system as self-organising? In [Ger07], it is argued that self-organising systems (in the sense of distributed systems) will have advantages in dynamic and unpredictable environments, where problems have to be solved that are not known beforehand and/or the addressed problem changes constantly. Then, a solution dynamically arises by local interactions and adaptation to unforeseen disturbances quickly appears. In theory, a centralised approach is also able to solve the problem, but in practice such an approach may require too much computation time to cope with the unpredictable disturbances in the system and its environment, e. g., when a system or its environment changes in less time than the system requires to compute a solution. In [Ed08], another brief sentence has been worked out to explain the main idea of self-organisation, being similar to the definition given by Carlos Gershenson in [Ger07]. A self-organising system consists of a set of entities that obtains an emerging global system behaviour via local interactions without centralised control. 3.1.4 Summary The investigation of self-organisation in many different disciplines of science has advantages and disadvantages at the same time. Many definitions from different domains have blurred the whole idea, which definitely is a disadvantage in terms of definition and terminology. On the other hand, the many disciplines keep the 20
- 53. 3.2 Emergence potential for many ideas and new approaches for creating (controlled) self-organising systems. This possibility will be even more attractive, if the research on self-organising systems converges towards a more standardised nomenclature, probably even forming a new field of science some day. Several positive effects from the interdisciplinarity of self-organisation became apparent when discussing possible ways to design the behaviour of the particular entities that form a self-organising system. The local behaviour is an integral part of a self-organising system, since the whole behaviour of the system emerges from the local interactions of the entities. In [Ed08], three basic approaches have been identified for finding a suitable set of local rules: Nature-inspired design, trial and error, and learning from an omniscient solution. In the case of dynamic self-organising systems, with distributed control and local interactions, emergence appears to be some kind of structure on a higher level. Since literature investigating self-organisation is also linked to the term of emergence, this phenomenon is addressed in the next section. 3.2 Emergence The phenomenon of emergence has been a fascinating topic for scientists such as John Stuart Mill [Mil43], George Henry Lewes [Lew75], and Conwy Lloyd Morgan [Llo23] for a long time, and the philosophical discussion of this topic is more than 150 years old. These so-called proto-emergentists consider the emergent process as a black box, where only the inputs and the outputs at the lowest level can be discerned without any knowledge about how the inputs are transformed into outputs. However, in the case of designing technical OC systems more recently characterised aspects of emergence need to be considered. A different perspective, referred to as neo-emergentism, summarises approaches of Jochen Fromm [Fro04, Fro05], John H. Holland [Hol98], Stuart Kauffman [Kau93], Aleš Kubík [Kub03], and others, where the root of emergence bases on the dynamics of a system, where investigations focus on reproducing the process, which leads to emergence, and where emergent phenomena are less miraculous than in the black box view. Emergence is the phenomenon occurring when a population of interconnected relatively simple entities self-organises to form more ordered higher level behaviour [Joh01]. Emergence can be referred to as the effect that the whole is greater than the sum of its parts. Emergent phenomena are defined by 1. the interaction of mostly large numbers of individuals 2. without centralised control with the result of 3. a global system behaviour, which has not explicitly been programmed into the individuals [Bea03]. 21
- 54. Chapter 3 Controlled Self-Organisation The journal Emergence2 , a journal of complexity issues in organisation and man- agement, provides the following characterisation of emergent behaviour. The idea of emergence is used to indicate the developing of patterns, structures, or properties that do not adequately seem explained by referring only to the system’s pre-existing components and their interaction. Emer- gence becomes of increasing importance, when the system is characterised by the following features. • When the organisation of the the system, i. e., its global order, appears to be more salient and of a different kind than the components alone; • when the components can be replaced without an accompanying decommissioning of the whole system; • when the new global patterns or properties are radically novel with respect to the pre-existing components; thus, the emergent patterns seem to be unpredictable and non-deducible from the components as well as irreducible to those components. Good examples for emergence originate from the observation of ants and other insects. The social insect metaphor for solving problems has become a diverse topic during the last years [BDT99]. For example, foraging behaviour in ants is characterised by the distribution of pheromones, thereby encouraging (but not forcing) other ants to follow the paths. This behaviour, despite its simplicity and distributedness, results in a very robust and efficient emergent phenomenon, i. e., that ants collectively find the shortest path between nest and food source. This observation has resulted in powerful metaheuristics for solving complex problems, called ant colony optimisation. Another example for emergent behaviour is the (human) brain. Although the exact function and interrelation of the different brain sub-systems is not really understood, scientists assume underlying emergent effects, as explained in [Rot05]. Today’s neurobiology is able to investigate those processes in human and animal brains in detail, which are responsible for the higher level cognitive functions like object recognition, attention, memorising, thinking, problem solving, action planning, empathy, and self-reflection, i. e., processes usually related to consciousness. It shows that these functions can uniquely be mapped to certain brain regions, and vice versa. This does not mean a violation of known physical/chemical/physiological laws. Neither are there any unexplainable gaps. Therefore, it seems necessary to view these brain functions as emergent states of a physical system. This is not in 2 http://www.emergence.org 22
- 55. Another Random Document on Scribd Without Any Related Topics
- 56. The story of how coal was first discovered in Belgium has been told a thousand times, but rarely, I think, in America. It seems that in a village not far from Liège there lived—some seven hundred years ago —a poor blacksmith named Houllos. One day he found himself quite out of money. He could not work to earn more, because he had no wood to heat his forge. While he sat bewailing his fate a mysterious stranger appeared and asked the cause of his woe. When he had heard the mournful story, “Take a large sack,” said he, “and go to the Mountain of the People. There you must dig down three feet into the earth. You will find a black, rocky substance, which you must put into the sack and bring home. Break it up, and burn it in your forge.” This is the reason why, in Belgium, coal still bears the name of huille, in memory of the blacksmith of Liège. Some think the stranger was an Englishman, since coal was already in use in London. But tradition has insisted that ange and not Anglais, is the proper word, and that Houllos entertained an angel. Near Mons are the great mounds of slag which were begun in the earliest times and look today not unlike the pyramids of Egypt. Whatever the origin of the mining industry in Belgium, there is nothing idyllic about the conditions there in modern times. The coal region of the Borinage is known as Le Pays Noir, and it certainly deserves the name. The miners are called Borains, or coal-borers. “They live both on the earth and in the earth, delving amid the black deposits of vast primeval forests.” Owing to their former long hours, which have been somewhat shortened in late years, the present generation is dwarfish, the men often under five feet and the women still less. Most of them cannot read or write, and they have little pleasure save what comes from beer. (More beer was sold per head in Belgium than even in Germany.) Of the hundred and twenty-five thousand miners in the country, three-quarters belonged to Hainault. There are in all over a hundred coal mines in Belgium, the area of those that were worked amounting to over ninety thousand acres, and of those not worked to forty thousand more. A new coal field has been discovered in the north but has not been exploited as yet. Although
- 57. the home consumption was steadily increasing, and averaged nearly three tons per capita, large amounts were exported to France and Holland. It was sold at a closer margin than in any other of the mining countries. Mining was commenced on the out-crops eight or nine hundred years ago, but it was only when steam-engines were invented that the miners were able to reach the deeper parts of the coal measures, and the yield was greatly increased. Firearms have been manufactured in Liège since midway in the fourteenth century. The first portable arms were the cannon and handgun, both adjusted to very heavy, straight butt-ends and very difficult to handle. They were loaded with stones, lead or iron balls. The musket and arquebus came later, and had matchlocks, an idea suggested by the trigger of the crossbow. The first exporters of Liège arms were naildealers, who possessed from immemorial times commercial relations with the most distant countries. After the invention of the flint-lock in the seventeenth century the gun trade made rapid progress. The number of workmen became enormous. The superiority of Liège arms was recognized all over the world, and the gunworkmen received offers of high salaries to induce them to go to France, England, Germany and Austria. Several of them were engaged to work at the Royal Manufactory of Arms at Potsdam. Much of the best work was done at the worker’s own house, and in order to prevent any decline in the individual skill of the men to whom Liège owed so much of its fame, the union of manufacturers of arms created a professional school of gunnery, where they could be specially trained. In this way they hoped to avoid the danger that the facility which machinery gives the workman would cause him to lose interest in his hand-work at home, which requires such varied knowledge and ability. Cotton spinning was one of the most important textile industries. Over a million spindles were employed, most of them in the two provinces of Hainault and Brabant, and in the city of Ghent. Most of the cotton came from America and Egypt.
- 58. An Old Lacemaker Verviers, in Liège, was the center of the wool-spinning industry. Here again the superior skill of the artisans established the reputation of the Belgian article. Most of the wool came from Australia and the Cape. For its flax spindles, however, Belgium raised its own material. The flax of Courtrai was considered the best in all Europe. More than half the finished thread was exported to England. The abundance of this material doubtless led to the early development of lace-making, for which the women of the country became so famous.
- 59. Flanders claims to be the birthplace of pillow-lace—dentelles aux fuseaux—and disputes with Italy the invention of lace generally. In earlier times drawn or cut work was often confused with lace, as was embroidery of one sort or another, and for this reason it is difficult to trace the art definitely back to its beginning. Ornamental needlework was done in Old Testament days, for Isaiah mentions those who “work in fine flax and weave networks.” But real lace-making—the interweaving of fine threads of flax, cotton, silk, of silver, gold or hair, to form a network—did not appear till the time of the Renaissance, when all the arts of Europe awoke to life. In a chapel at St. Peter’s, in Louvain, was an altar-piece painted in 1495 by Quentin Matsys, which showed a girl making lace on a pillow like those still in use to this day. The manufacture of lace began in Brussels about the year 1400. The city excelled from the first in the quality of the work done there. This was due to the fineness of the thread of Brabant, which the women spun inch by inch with such painstaking care that it defied competition. A pound of flax was sometimes transmuted into lace worth several thousand dollars. The lace industry was the only one in Flanders which survived the upheavals of the sixteenth century. Its prosperity alone tided the distracted people over their difficulties and saved them from the ruin which threatened. The women plodded on at their slow task, hour after hour, thread after thread, for a pitiful few cents a day, and never knew that they had saved their country. “They are generally almost blind before thirty years of age,” wrote an early chronicler. The women of Belgium have always been specially adept with the needle, and it may be that the rainy weather so prevalent there had something to do with the development of this indoor industry. Certainly lace-making is—or was, until very recently—practised in all the provinces except Liège, and in some districts it could be said that every woman, young or old, handled the bobbins or the needle. It was, indeed, the national industry. As a rule, the women worked to order and by contract, and were paid by the piece. The lace, when finished, was handed over to the local
- 60. middleman, who, in turn, sold it to the contractors in the cities. The children learned the art from their mother or—more often—from the nuns in the various convent schools. They would enter these schools when six or eight years old, and often remained there till their marriage. The nuns did much to keep up the ancient traditions of the art, and even in their convents in the Far East today they make a point of teaching the native children to copy European laces. There are two kinds of lace, point and pillow. The former is made with a needle, and its characteristic feature is the “set-off” of the flowers. The needle laces, of Belgium are divided into Brussels point, Brussels appliqué, Venice, rose and Burano points. Several classes of workers are needed for each piece—those who make the openwork ornaments and the flowers, and those who apply them on to the background, a very delicate task. Brussels point is the finest example of this form of lace, and indeed of any lace made in Belgium at the present time. The designs are very elaborate, with the flowers often in relief. Modern Brussels point is, however, too frequently an imitation, with flowers sewn on to a machine-made net that is often rather coarse, while the application is done by unskilled fingers. Of pillow lace there are many kinds, and their chief characteristic is the outline of the design. The lace is made on a cushion or pillow which stands on a frame, with little spools or bobbins for the threads, and pins for fixing the lace on the pattern. The best kinds of pillow lace are duchess, Mechlin, and Valenciennes. “Valenciennes the eternal,” they called it, because by working fourteen hours a day for a year you made less than half a yard. Marie Therèse had a dress of it which took a year to make and cost fourteen thousand dollars. Considering that the workers received barely a cent an hour, one gets some idea of the magnitude of the task. The Béguinage in Ghent was the headquarters for the manufacture of this lace, but only a few old nuns remain there now who know the secrets of its making. Machine-made imitations flood the market, and the
- 61. former process is too costly to make it worth any one’s while to master it. BRUSSELS POINT LACE. Mechlin is the Flemish name for the town of Malines, and both words are used in connection with the lace which originated there. Mechlin is the airiest and most exquisite of laces, but its very delicacy made it too costly, and since it could be so easily and cheaply imitated, it is no longer made by hand. It was constructed in one piece, with no application, a flat thread forming the flower and giving it almost the
- 62. appearance of embroidery. Napoleon, who admired it greatly, cried out when he saw the delicate spire of Antwerp Cathedral that it was like “la dentelle de Malines.” In spite of the fact that the art of making lace had fallen upon hard days, the lacemakers’ ball was still an important event of the season when we were in Brussels. It came in carnival week, and was the occasion on which the Société de la Grande Harmonie received the King and Queen. It interested me to see how Their Majesties were welcomed by such a representative body of middle-class citizens— there was the most genuine enthusiasm I have ever seen shown towards royalty. The Diplomatic Corps had been invited to attend, and we were taken to a platform at the end of a great room, where the royal chairs were placed, and chairs in rows for the Corps and the Court and the Ministers of State. Beyond the columns which divided the hall into three parts were arranged the seats for the members of the society. The center of the floor remained clear, and here the tableaux and pageants representing the various stages in the history of lace were performed. In their pageant the lacemakers all wore examples of their craft. One of the prettiest incidents occurred when the groups of costumed personages separated and there passed along the length of the ballroom floor two little children, a boy and a girl, dressed as a page and a miniature lady-in-waiting. They advanced slowly, and presented to the King and Queen books which told of the evening’s entertainment. The Queen rose and apparently questioned the president of the society about the little girl who stood so shyly before her. Then, taking the book, she stooped down and kissed her. It was very prettily and naturally done, and caused a round of appreciative applause and cries of “Long live the Queen!” Another attractive feature was that of the tiny children who represented the Flemish lacemakers, each one wearing the costume of the trade. They passed in procession before the Queen and each, with a little courtesy, laid a bouquet of flowers at her feet.
- 63. I was surprised to find that Brussels was the market for lace from all over the world, and that foreign laces of every description were copied there by the skilful dentellières. This was still true, in spite of the marked decline which the industry had shown of late, especially since the introduction of machinery. Where a generation ago one hundred and fifty thousand women were employed, in 1910 there were barely twenty thousand. Their product had lost in quality, too, as well as in quantity. The old nuns who did the wonderful, intricate stitches, were dying off and there were none to take their places. The pattern-makers, also, were contenting themselves with easier designs. Belgium was “speeding up,” with the rest of the world, and the painstaking arts had to suffer. Modern laces are carelessly made, in comparison with those of former days, and from inferior designs. The wages paid those who still work at the craft seem low indeed, especially when the long years of apprenticeship are considered. Verhaegen, in statistics collected in 1910, cites a girl of thirteen who was working ten hours a day, making in fifty-five hours a meter of Cluny lace for which she received about fifty cents. Children of fourteen were working seventy-two hours a week for something less than a cent an hour, and grown women earned little more. The workers were not organized, and the middlemen seem to have prospered accordingly. But the pay was low in all branches of industry, even those which were well organized. An English writer noted that the rate of wages per hour for men in Belgium was only about half that prevailing in Britain, while the cost of living was nearly the same. The average earnings of the breadwinner of the family were about $165 a year. These facts certainly account for the development of coöperation. This movement, which had a great vogue throughout the country, started in Ghent in 1873. Bread was scarce, and famine prices prevailed. A group of poor weavers conceived the idea of baking for themselves and their friends at cost. Their capital consisted of the vast sum of seventeen dollars and eighteen cents. Their bakery was in a
- 64. cellar, and their utensils were antiquated. They could not afford a dog to deliver their wares, which were taken from door to door in a basket. But this was only the beginning. The “free bakers,” as they called themselves, came to have for their headquarters one of the finest buildings in Ghent. A few years later Edouard Anseele, realizing the power of the new movement, decided that it should be identified with Socialism for their mutual benefit. To that end was organized the Vooruit, which has branches all over Belgium, and in other countries as well. Instead of returning the profits made on bread sold at market prices to the purchasers, as had been originally done, a percentage was retained for the support of the organization in its various departments. There was a mutual benefit fund, for instance: bread was sent to members out of work; a doctor went to those who were ill; a trained nurse was at hand to look after the first baby and to instruct the mother in its care. When the Church set up rival bakeries, the Vooruit went farther. It established its first “maison du peuple,” which has since been duplicated in many places. Every need of the people was supposed to find here its satisfaction. There was a café, with tables in the park, and lights and music. There were lectures, dances, debates, concerts, movies. There was a theater where the actors and the plays were chosen by the vote of the audience, which, by the way, strongly favoured their own Maeterlinck. Besides a library and a day nursery, there was a big department store, and in the same building were the headquarters for all the allied and friendly organizations—trade unions, coöperative and socialistic societies, and so on. One of the most interesting activities of the Vooruit was the traveling club for children, bands of whom went from town to town, picking up recruits as they went, seeing their own land first, then—this was before the war—crossing the border into France or Germany, where the local Vooruits made them welcome. A common practice was for children of the French and Flemish parts of the country to be
- 65. exchanged for long visits, so that they might have a chance to learn each other’s language. When the organization, which had always before refused to sell alcoholic drinks, found itself bitterly opposed by the liquor interests, especially in the mining districts, it built breweries of its own. In this way it was able to give the working men pure beer at a very low cost. The Maison du Peuple in Brussels was established in 1881, with a capital of about one hundred dollars. It began, like the one in Ghent, as a bakery, and owned a dog and a small cart to make deliveries. At last accounts the society had over ninety dogs. It is amusing to read that these had their own kitchens, where their cooking was done, and their bathrooms, where they were kept clean. And when one is speaking of the workers of Belgium, the dogs should not be forgotten, for the larger breeds were very useful members of the industrial system. Laundresses, bakers and vendors used them in distributing their wares, and they were of great service on the farm. But perhaps the commonest sight was that of a dog hitched to a cart filled with shining brass and copper milk cans. They were all carefully inspected to see that their harness fitted properly, and that they were provided with a drinking bowl and with a mat to lie down on when they were tired. The Government made a point, indeed, of seeing that conditions were as comfortable as possible for the animals. The poor cannot afford to keep a dog simply for a pet; there are no scraps from the table to feed him, because no thrifty housewife leaves any scraps; he must do his share and earn his keep like the others. At a time when France laid a heavy tax on imported laces, dogs made excellent smugglers. They were kept for a time on the French side of the line, petted and well fed; then they were sent over into Belgium, where they were allowed to become thoroughly homesick. Skins of larger dogs were lined with contraband lace and tied on to them, and they were headed for home and set free. Of course they naturally sought their own firesides, and the lace went with them. When the
- 66. ruse was discovered, over forty thousand of them were captured and put to death. Since the war began, dogs have been of great service in dragging the mitrailleuses, the light machine-guns, as well as in helping their masters carry their household goods to a place of safety. The police dogs were wonderfully trained, and have been used by the Red Cross to find the wounded in remote places and to carry first aid. The same high standards of efficiency by which Belgian workmen made a national reputation for their various manufactures showed also in the cultivation of the ground. The whole western part of the country was one vast market-garden, but it was no happy chance of soil and climate that made it so. Generations of unbroken toil on the part of a patient, skilful peasantry, equipped with the most primitive tools but with a positive genius for their work, were necessary. So recently as the first half of the nineteenth century there was a wild stretch of land west of the Scheldt known as the Pays de Waes, which was uncultivated and desolate. Today it is wonderfully fertile, its little truck farms supporting five hundred people to the mile.
- 67. "SINCE THE WAR BEGAN, DOGS HAVE BEEN OF GREAT SERVICE IN DRAGGING THE MITRAILLEUSES." Flanders as a whole, indeed, had poor soil, often “an almost hopeless blowing sand.” The method of reclamation usually began with the planting of oats, rye or broom. This was used three years for forage and then plowed in, after which the land became capable of producing clover. The rotation of crops was worked out with great care, according to the special needs of the soil. The Belgian wheat crop averaged thirty-seven bushels to the acre in 1913, while in the same year “up-to-the-minute” America raised only fifteen bushels. The soil is particularly suited to hemp and flax, the latter furnishing not only oil but fiber, of which the British markets bought ten million dollars’ worth annually. Poppies were grown for oil. Tobacco yielded two tons to the acre, and white carrots eight hundred bushels. The Flemish farmer did most of his work by hand, with no other implement than a spade, which has been called the national tool. The population was so large that human labour was cheaper than animal. In sixteen days a man could dig up an acre of land as well as a horse could plow it. A farmer was able to support himself, his wife and three children, keep a cow and fatten a hog, on two and a half acres. With another acre he had a surplus product to carry to market. A man with a capable wife and children could do all the work on six acres and have time left for outside interests. If he was fortunate enough to have horses they were the pride of his heart and he kept them always finely groomed and in the pink of condition. The women of the country married early, raised large families, and worked hard. They were good managers, especially in the Walloon districts where they often carried on some industry besides their housekeeping. For centuries their chief employment was making lace. The Government established schools of housekeeping, where the girls learned domestic economy in every branch; they were sent to market,
- 68. for instance, with six cents to buy the materials for a meal, which they afterwards cooked and served. The Government indeed did everything it could to improve conditions in the country districts and to encourage farming. It established schools of agriculture, with dairy classes for the girls, and aided in starting coöperative societies. Its policies were far-seeing and marked by a really paternal interest, as well they might have been, for to her sturdy peasants—and to the peasants’ sturdy wives—were due the foundations of Belgian prosperity.
- 69. CHAPTER IX TAPESTRIES S we were intensely interested in tapestries we often went to the Museum to study and admire the most famous set in Brussels, an early Renaissance series of four pieces, called Notre Dame du Sablon. These hangings illustrate an old fourteenth-century story, which I condense from Hunter’s delightful work on “Tapestries.” Beatrix Stoelkens, a poor woman of Antwerp, was told by the Virgin in a dream to get from the church of Notre Dame a little image of the Madonna. In obedience to the vision she obtained the statuette and took it to a painter, who decorated it in gold and colours. After Beatrix had returned it to the church, the Virgin clothed it with such grace that it inspired devotion in all who saw it. Then Our Lady appeared a second time to Beatrix, and directed her to carry the statue to Brussels. When she attempted to get it, the warden of the church interfered, but he found himself unable to move, and Beatrix bore away the little Madonna in triumph. She embarked for Brussels in an empty boat, which stemmed the current as if piloted by unseen hands. On arriving at her destination, she was received by the Duke of Brabant and the magistrates of the city, and the precious little statue was carried in procession to the church of Notre Dame du Sablon. This set bears the date 1518, when Brussels was no longer under a Burgundian Duke, but Charles V was ruler of the Netherlands. The designer of the set followed the Gothic custom of representing the story under the forms of his own day, so, instead of the Duke of Brabant, Philip the Fair, father of Charles V, is pictured receiving the Madonna from the hands of Beatrix at the wharf, Charles V and his brother Ferdinand are bearing it in a litter to the church, and Margaret
- 70. of Austria, aunt of Charles, kneels in prayer before the niche where the sacred image has been placed. When in New York it always gives us pleasure to go to the Metropolitan Museum to see the finest Belgian set in the United States, the Burgundian Sacraments, woven in the early fifteenth century. This splendid example of Gothic workmanship was made in the days when Philip the Good had brought the power of Burgundy to its zenith. When the great Duke wanted to have magnificent hangings for the chamber of his son (who was afterward Charles the Bold), he ordered a set of tapestries from the weavers of Bruges. All that remains of this splendid work of art is now in the New York Museum— five pieces, which form half of the original set. The complete series consisted of two rows of scenes, the upper seven representing the Origin of the Seven Sacraments, the lower, the Seven Sacraments as Celebrated in the Fifteenth Century. This set shows wonderful weaving, “with long hatchings that interpret marvelously the elaborately figured costumes and damask ground.” There are other exquisite tapestries in America, too, for the Committee of Safety in 1793 imported some American wheat into France, and when the time came to pay it proffered assignats. Naturally enough, the Americans objected, but there was no money. “Then they offered, and the United States was obliged to accept in payment, some Beauvais tapestries and some copies of the Moniteur.” Tapestries required muscular strength, for the material was heavy, and so men were given this work in town workshops. The ladies did the needle, bobbin and pillow work in the castles and convents. True tapestry is always woven on a loom, and is a combination of artistic design with skill in weaving. This tapestry industry was introduced into Western Europe in the Middle Ages by the Moors, but we can trace the art of making woven pictures to much earlier times. The ancient Romans had them. Ovid describes the contest in weaving between Arachne and Pallas, in which the maiden wrought more beautifully than the goddess. Pallas in anger struck the maid, who hanged herself in her rage because she
- 71. dared not return the blow. The goddess, relenting, changed Arachne into a spider, and she continues her weaving to this day. But a much earlier poet has described the making of tapestry. We read in the Odyssey that, when the return of Ulysses to his native land was long delayed, his faithful wife Penelope postponed a decision among the suitors who importuned her by promising to make a choice when she had finished weaving the funeral robe for Laertes, her husband’s father. The robe was never completed, for each night she took out the work of the day before. It is a very interesting fact that a Grecian vase has come down to us on which is a painting of Penelope and her son Telemachus. Penelope is seated at what the experts say is certainly a tapestry loom, though somewhat different from those used at a later day. We have no large pieces done by the Greeks and the Romans, but many small bands for use as trimmings of robes. Some of these were woven by the Greeks as early as the fourth century B.C., others were made in Egypt under Roman rule some centuries later, and are called Coptic. From these one can trace the series through the silken Byzantine, Saracenic and Moorish dress tapestries to the Gothic fabrics of the fourteenth century. The Flemish and Burgundian looms were those of Arras, Brussels, Tournai, Bruges, Enghien, Oudenarde, Middlebourg, Lille, Antwerp, and Delft in Holland. The value of the tapestry industry to Flanders may be judged from the fact that Arras, a city of no importance whatever, from which not a single great artist had come, led all Europe for about two centuries in tapestry weaving. Although some fine pieces were woven in the fourteenth century, as far as known, only two sets of Arras tapestries of this period are left. One set is at the cathedral of Angers in rather bad condition, for they were not appreciated at one time, and were used in a greenhouse and cut up as rugs. Fortunately, they have been restored and returned to the cathedral. The other set of early Arras hangings is to be found at the cathedral of Tournai, in Belgium. A piece of this set bore an
- 72. inscription—which has fortunately been preserved for us—stating, “These cloths were made and completed in Arras by Pierrot Féré in the year one thousand four hundred two, in December, gracious month. Will all the saints kindly pray to God for the soul of Toussaint Prier?” Toussaint Prier, a canon of the cathedral in 1402, was the donor of the tapestries. When Louis XI of France captured Arras, in 1477, and dispersed the weavers, Tournai, Brussels, Oudenarde and Enghien took up the work. The oldest Brussels tapestries known belong to the latter part of the fifteenth century. Two of these sets were painted by Roger van der Weyden and celebrated the Justice of Trajan and the Communion of Herkenbald. Some have tried to prove that other important tapestries were designed by the great primitives, but Max Rooses assures us the resemblance to their work comes from the fact that their characteristics, “careful execution, extreme delicacy of workmanship, and brilliancy of colour,” pervaded every branch of art at that period. Brussels and Oudenarde held the lead throughout the sixteenth century. The Bruxellois wove vast historical compositions to decorate the palaces of kings; the weavers of Oudenarde produced landscapes, “verdures” and scenes from peasant life for the homes of burghers. Tapestries are at their best as line drawings; when more complicated effects are sought “confusion and uncertainty follow.” The finest ever woven were produced during the last half of the fifteenth and the first half of the sixteenth centuries, when Gothic tapestries gradually ceased to be made and Renaissance pieces began to take their place. During that hundred years, when the weavers were most skilful and were still satisfied with line drawings, many of the finest tapestries combined the characteristics of both styles. In the sixteenth century, the weavers had such marvelous skill, however, that they actually reproduced the shadow effects of Italian designs. Even such great artists as Raphael and Michael Angelo drew cartoons, and stories of ten, twenty or even thirty scenes were woven, all showing the distinctive characters of Renaissance art. They combined breadth of composition and lively action with the
- 73. introduction of nude figures and elaborate landscape and architectural settings. But in trying to copy painting too closely, they departed from the best traditions of tapestry technique, and deterioration was sure to follow in time. After the desolating wars of the sixteenth century, when arts and industries revived under the Archdukes Albert and Isabella, Brussels weavers set up their looms again, and “Rubens brought new life into tapestry manufacture. He supplied the Brussels workshops with four great series—the History of Decius Mus, destined for some Genoese merchants; the Triumphs and Types of the Eucharist, ordered by the Infanta Isabella for the convent of the Clares at Madrid; the History of the Emperor Constantine, executed for Louis XIII; and the History of Achilles, for Charles I.... The Triumphs and Types of the Eucharist are the most powerful allegories ever created to glorify the mysteries of the Catholic religion.”[4] Jacob Jordaens also designed tapestry cartoons, but the most popular artist among the weavers at the end of the seventeenth and in the eighteenth centuries was David Teniers. He did not himself make designs, but the manufacturers, especially at Oudenarde, borrowed his subjects, which were drawn largely from peasant and village life. One reason why we have so few of the really antique tapestries is that in 1797 the market for them was so dead—owing to the increasing use of wall-papers and canvases painted in oils—that the French decided it would be better to burn them for the gold and silver they contained. Accordingly, “One hundred and ninety were burned. During the French Revolution, a number of tapestries that bore feudal emblems were also burned at the foot of the Tree of Liberty.” At this time, when they were not in fashion, many rare old hangings were cut up by the inartistic or the ignorant and used as rugs and curtains. But in recent years, we are told, the Brothers Braquenié have set up a workshop at Malines, where they have produced a fine series for the Hôtel de Ville in Brussels, called “Les Serments et les Métiers de Bruxelles.” The cartoons for this set were made by Willem Geefs, the painter.
- 74. As to the material, there is a great difference. Gothic tapestries are composed of woolen weft on linen, or woolen on hemp warp, and are often enriched with gold and silver thread. These are not used today, as they are considered too expensive. Since the sixteenth century, Brussels, Gobelins, and Mortlake have used a great deal of silk. In the fifteenth century fifteen or twenty colours were employed, in the Renaissance period, twenty or thirty. “Both high warp and low warp antedated the shuttle. In other words, they use bobbins that travel only part way across instead of shuttles that travel all the way across.” The high warp loom was also in use before the treadle. “In the low warp loom the odd threads of the warp are attached to a treadle worked with the left foot, the even threads of the warp to a treadle worked with the right foot, thus making possible the manipulation of the warp with the feet and leaving both hands free to pass the bobbins. In the high warp loom, that has no treadle, the warps are manipulated with the left hand while the right hand passes the bobbins back and forth. The term high warp means that the warp is strung vertically, low warp horizontally.” Both are woven with the wrong side toward the weaver. “The wrong side in all real tapestries is just the same as the right side except for reversal of direction and for the loose threads.... In the high warp loom, the outline of the design is traced on the warp threads with India ink from tracing paper, and the coloured cartoon hangs behind the weaver, where he consults it constantly. In the low warp loom, the coloured cartoon is usually beneath the warp, and often rolls up with the tapestry as it is completed.”[5] In the eighteenth century, the low warp loom was considered better than the haute lisse, or high warp. Great care has to be taken in dyeing the threads of the weft, which are much finer than those of the warp. Vegetable dyes, such as cochineal, madder, indigo, etc., must be used, for permanent colours can never be obtained with aniline dyes. The old Spanish dyes were considered the best. In this country, one sometimes gets the fine colours in an old Mexican serape or a prized Navajo blanket. The wool that is used to mend old tapestries in the American museums is
- 75. coloured with dyes made by Miss Charlotte Pendleton in her workshop near Washington, which I have visited. The Arras tapestries have a better and more attractive texture than any others. “Arras tapestries are line drawings formed by the combination of horizontal ribs with vertical weft threads and hatchings. There are no diagonal or irregular or floating threads, as in embroideries and brocades. Nor do any of the warp threads show, as in twills and damasks. The surface consists entirely of fine weft threads that completely interlace the coarser warp threads in plain weave (over and under alternately), and also completely cover them, so that only the ribs mark their position—one rib for each warp thread. Every Arras tapestry is a rep fabric, the number of ribs eight to twenty-four to the inch.” The finely woven textures are not always considered the best. “The most marvelous tapestries of the fifteenth century were comparatively coarse (from eight to twelve ribs), and of the sixteenth were moderately coarse (from ten to sixteen).” Many of the early Gothic tapestries had inscriptions woven at the bottom or the top, but had no borders. It was not until toward the end of the fifteenth century that they began to develop these. They first had narrow verdure edgings, until Raphael introduced compartment borders in the set of the Gates of the Apostles, the most famous tapestries of the world. The most noted cartoons in existence are the designs for this set, in the Victoria and Albert Museum at South Kensington. Renaissance borders were much wider than the Gothic, and were filled with greens and flowers. At the end of the seventeenth century the borders took the form of imitation picture frames. Gothic verdures are in reality coloured drawings in flat outline of trees and flowers with birds and animals. Renaissance verdures have more heavily shaded leaves and look more true to nature. The majority of Gothic tapestries are anonymous as regards both maker and designer. With the Renaissance began the custom in Brussels and other Flemish cities of weaving the mark of the city into the bottom selvage, and the monogram of the weaver into the side selvage on the right. This custom was established by a city ordinance
- 76. of Brussels in 1528. An edict of Charles V made it uniform, in 1544, for the whole of the Netherlands. After another century, weavers began to sign their full names or their initials in Roman letters, and monograms were discarded. When the weavers of Arras took refuge in other countries, after the capture of that town by Louis XI, they went by thousands to England and France. In this way the French looms at Gobelins, Beauvais, and Aubusson were started, and those at Mortlake, in England. As early as the fourteenth century, there was at least one eminent master weaver in Paris, Nicolas Bataille, in whose factory part of the remarkable Apocalypse set of the cathedral of Angers was woven. But even in the fifteenth and sixteenth centuries, French tapestries were far from equaling those of Flanders. In 1667, Colbert “established in the buildings of the Gobelins the furniture factory of the Crown under the direction of Charles Lebrun.” The great establishment of “Les Gobelins,” by the way, has an interesting history. Jean and Philibert Gobelin built a dyehouse in the fifteenth century by the little stream of the Bièvre, in the Faubourg, whose waters had peculiar qualities that gave special excellence to their dyes. The family found dyeing so profitable that they were able to become bankers, and at the beginning of the seventeenth century they sold the establishment, which, however, still kept their name. Here Comans and Planche, tapestry weavers from Flanders, opened a factory in 1601. The edict of Henri Quatre by which they were incorporated gave them important privileges, but also obliged them to train apprentices and to establish the craft in the provinces. During the Spanish occupation of the Netherlands, many tapestries were taken to Spain, where the finest in existence today are to be found. They may be seen in the churches and draping the balconies over the streets of a fête day. King Alfonso owns seven miles of gold and silver thread hangings. But these are only the remnant of what Spanish royalty formerly possessed. Charles V, Philip II, and many others of the ruling house were indefatigable collectors. The famous Conquest of Tunis, in twelve pieces, was woven by Willem de
- 77. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com