Ramos, v; fernandes, c; rosa, a. (2007): social cognitive maps, swarm collective perception and distributed search on dynamic landscapes

Social Cognitive Maps, Swarm Collective Perception
and Distributed Search on Dynamic Landscapes
Vitorino Ramos, Carlos Fernandes, Agostinho C. Rosa
LaSEEB-ISR-IST, Evolutionary Systems and Biomedical Eng. Lab.,
Technical Univ. of Lisbon (IST),
Av. Rovisco Pais, 1, TN 6.21, 1049-001, Lisbon, PORTUGAL
{vramos, cfernandes, acrosa}@laseeb.org

Abstract. Swarm Intelligence (SI) is the property of a system whereby the collective

behaviors of (unsophisticated) entities interacting locally with their environment cause coherent
functional global patterns to emerge. SI provides a basis with which it is possible to explore
collective (or distributed) problem solving without centralized control or the provision of a
global model. To tackle the formation of a coherent social collective intelligence from
individual behaviors, we discuss several concepts related to self-organization, stigmergy and
social foraging in animals. Then, in a more abstract level we suggest and stress the role played
not only by the environmental media as a driving force for societal learning, as well as by
positive and negative feedbacks produced by the many interactions among agents. Finally,
presenting a simple model based on the above features, we will address the collective
adaptation of a social community to a cultural (environmental, contextual) or media
informational dynamical landscape, represented here – for the purpose of different experiments
– by several three-dimensional mathematical functions that suddenly change over time. Results
indicate that the collective intelligence is able to cope and quickly adapt to unforeseen
situations even when over the same cooperative foraging period, the community is requested to
deal with two different and contradictory purposes.

1 Introduction
Swarm Intelligence (SI) is the property of a system whereby the collective behaviors
of (unsophisticated) entities interacting locally with their environment cause coherent
functional global patterns to emerge. SI provides a basis with which it is possible to
explore collective (or distributed) problem solving without centralized control or the
provision of a global model (Stan Franklin, Coordination without Communication,
talk at Memphis Univ., USA, 1996). The well-know bio-inspired computational
paradigms know as ACO (Ant Colony Optimization algorithm [8]) based on trail
formation via pheromone deposition / evaporation, and PSO (Particle Swarm
Optimization [24]) are just two among many successful examples. Yet, and in what
specifically relates to the biomimicry of these and other computational models, much
more can be of useful employ, namely the social foraging behavior theories of many
species, which can provide us with consistent hints to algorithmic approaches for the
construction of social cognitive maps, self-organization, coherent swarm perception
and intelligent distributed search, with direct applications in a high variety of social
sciences and engineering fields. In the present work, we will address the collective

1 Ramos V., Fernandes C., Rosa A.C., submitted to Advancing Artificial
Intelligence through Biological Process Applications, IDEA Group, 2007.

adaptation of a social community to a cultural (environmental, contextual) or
informational dynamical landscape, represented here – for the purpose of different
experiments – by several three-dimensional mathematical functions that change over
time.
Flocks of migrating birds and schools of fish are familiar examples of spatial selforganized patterns formed by living organisms through social foraging. Such
aggregation patterns are observed not only in colonies of organisms as simple as
single-cell bacteria, as interesting as social insects like ants and termites as well as in
colonies of multi-cellular vertebrates as complex as birds and fish but also in human
societies [14]. Wasps, bees, ants and termites all make effective use of their
environment and resources by displaying collective “swarm” intelligence. For
example, termite colonies build nests with a complexity far beyond the comprehension
of the individual termite, while ant colonies dynamically allocate labor to various vital
tasks such as foraging or defense without any central decision-making ability [8,53].
Slime mould is another perfect example. These are very simple cellular organisms
with limited motile and sensory capabilities, but in times of food shortage they
aggregate to form a mobile slug capable of transporting the assembled individuals to a
new feeding area. Should food shortage persist, they then form into a fruiting body
that disperses their spores using the wind, thus ensuring the survival of the colony
[30,44,53].
New research suggests that microbial life can be even richer: highly social, intricately
networked, and teeming with interactions [47]. Bassler [3] and other researchers have
determined that bacteria communicate using molecules comparable to pheromones. By
tapping into this cell-to-cell network, microbes are able to collectively track changes
in their environment, conspire with their own species, build mutually beneficial
alliances with other types of bacteria, gain advantages over competitors, and
communicate with their hosts - the sort of collective strategizing typically ascribed to
bees, ants, and people, not to bacteria. Eshel Ben-Jacob [6] indicate that bacteria have
developed intricate communication capabilities (e.g. quorum-sensing, chemotactic
signalling and plasmid exchange) to cooperatively self-organize into highly structured
colonies with elevated environmental adaptability, proposing that they maintain
linguistic communication. Meaning-based communication permits colonial identity,
intentional behavior (e.g. pheromone-based courtship for mating), purposeful
alteration of colony structure (e.g. formation of fruiting bodies), decision-making (e.g.
to sporulate) and the recognition and identification of other colonies – features we
might begin to associate with a bacterial social intelligence. Such a social intelligence,
should it exist, would require going beyond communication to encompass unknown
additional intracellular processes to generate inheritable colonial memory and
commonly shared genomic context. Moreover, Eshel [5,4] argues that colonies of
bacteria are able to communicate and even alter their genetic makeup in response to
environmental challenges, asserting that the lowly bacteria colony is capable of
computing better than the best computers of our time, and attributes to them properties
of creativity, intelligence, and even self-awareness.
These self-organizing distributed capabilities were also found in plants. Peak and coworkers [37,2] point out that plants may regulate their uptake and loss of gases by
distributed computation – using information processing that involves communication


between many interacting units (their stomata). As described by Ball [2], leaves have
openings called stomata that open wide to let CO2 in, but close up to prevent precious
water vapour from escaping. Plants attempt to regulate their stomata to take in as
much CO2 as possible while losing the least amount of water. But they are limited in
how well they can do this: leaves are often divided into patches where the stomata are
either open or closed, which reduces the efficiency of CO2 uptake. By studying the
distributions of these patches of open and closed stomata in leaves of the cocklebur
plant, Peak et al. [37] found specific patterns reminiscent of distributed computing.
Patches of open or closed stomata sometimes move around a leaf at constant speed,
for example. What’s striking is that it is the same form of mechanism that is widely
thought to regulate how ants forage. The signals that each ant sends out to other ants,
by laying down chemical trails of pheromone, enable the ant community as a whole to
find the most abundant food sources. Wilson [54] showed that ants emit specific
pheromones and identified the chemicals, the glands that emitted them and even the
fixed action responses to each of the various pheromones. He found that pheromones
comprise a medium for communication among the ants, allowing fixed action
collaboration, the result of which is a group behaviour that is adaptive where the
individual’s behaviours are not.
Some other authors also defend this self-organizing realm into brain function. As
defended in [12], the self-organization of ants into a swarm and the self-organization
of neurons into a brain-like structure are similar in many respects. As described earlier
in here, swarms of social insects construct trails and networks of regular traffic via a
process of pheromone laying and following. These patterns constitute what is known
in brain science as a cognitive map. The main differences lies in the fact that insects
write their spatial memories in the environment, while the mammalian cognitive map
lies inside the brain, a detail that also constitutes an important advantage for artificial
cognitive map formation in digital realms presented in past models [39]. As mentioned
by the two authors, this analogy can be more than a poetic image, and can be further
justified by a direct comparison with the neural processes associated with the
construction of cognitive maps in the hippocampus. In [54], Wilson forecasted the
eventual appearance of what he called “a stochastic theory of mass behavior” and
asserted that “the reconstruction of mass behaviors from the behaviors of single
colony members is the central problem of insect sociobiology”. He forecasted that our
understanding of individual insect behavior together with the sophistication with
which we will able to analyze their collective interaction would advance to the point
were we would one day posses a detailed, even quantitative, understanding of how
individual “probability matrices” would lead to mass action at the level of the colony.
As stated in [12], by replacing colony members with neurons, mass behaviors or
colony by brain behavior and insect sociobiology with brain science the above
paragraph could describe the paradigm shifts in the last twenty-five years of progress
in the brain sciences [54] and Artificial Intelligence (AI) research. From traditional AI
to cognitive sciences and connectionist models. In contrast with symbol processing
and traditional AI approaches, in connectionist models representations of any
phenomena are dynamically stored in the matrix of weights between pairs of
processing elements and patterns to be retrieved are not explicitly stored anywhere.
Retrieval is accomplished by activating a pattern of inputs, which is processed by the


network of constituent elements, resulting in output of a recalled pattern. The answer
is somewhere spread in the dynamics and interactions among several units of the
system, not on the units itself, in conformity with Chris Langton’s Artificial Life
research area definition [26,40]. Either if these units are ants, bacteria, plants, fish,
cells of an immune system or neurons. As recently stated by Chialvo [13,48], highly
correlated brain dynamics produces synchronized states with no behavioral value,
while weakly correlated dynamics prevents information flow. In [13], Chialvo discuss
the idea put forward by Per Bak [1] that the working brain stays at an intermediate
(critical) regime characterized by power-law correlations.
As early as 1979, Douglas Hofstadter [19,24] suggested that maybe the brain was like
an ant colony. No single neuron in a brain contains knowledge in itself, it is only
through their interaction that thinking can occur. As Hofstadter’s Anteater explains:
There is some degree of communication among the ants, just enough to keep them
from wandering off completely at random. By this minimal communication they can
remind each other that they are not alone but are cooperating with teammates. It
takes a large number of ants, all reinforcing each other this way, to sustain any
activity – such as trail building – for any length of time. Now my very hazy
understanding of the operation of brain leads me to believe that something similar
pertains to the firing of neurons... (p. 316, [19]).
As acknowledged in [24], Hofstadter’s view of ants is in line with the contemporary
appreciation for the emergence of complex dynamical systems from the interactions of
simple elements following simple rules. Like Hofstadter, Millonas [33,32,12]
compares the communications network within a swarm of ants (pheromone-based and
extremely dynamic) to the highly interconnected architecture of neurons in a brain.
Both cases (nodes or units is in here an abstract term) can be described in terms of
three characteristics: (1) their structure comprises a set of nodes and their
interconnections, (2) the states of node variables change dynamically over time, and
(3) there is learning – changes in the strengths of the connections among the nodes. An
argument that goes in line with the famous Doyne Farmer’s 1991 connectionism paper
[16]. Millonas argues that the intelligence of an ant swarm arises during phase
transitions – the same phase transitions that Langton described as defining the “edge
of chaos” [27,26], or Wolfram’s Class IV state [55], identifying a special dynamical
regime – with open-ended novelty - squashed between chaotic regimes and the ones
characterized by fixed points and cycle limits. Langton have suggested that the critical
region between ordered and chaotic behaviour plays a central role in evolution, using
a simple numerical parameter λ for the measure of complexity in a CA (Cellular
Automata). Systems that are in a state lying close to this edge of chaos (a critical value
of λ) have the richest, most complex behaviour. This phase transition argument is also
synchronized with the Self-Organized Criticality (SOC) theory of Per Bak [1],
habitually recognized worldwide in his sand pile model, or in the frequency-size of
earthquakes as well as city areas following a typical 1/f noise decay curve. As stated in
[24], the movements of ants are essentially random as there is no systematic
pheromone pattern; activity is a function of two parameters, which are the strength of
the pheromone and the attractiveness of the pheromone to the ants. If the pheromone


distribution is random or if the attraction of ants to the pheromone is weak, then no
aggregation pattern will form. On the other hand, if a too-strong pheromone
concentration is established, or if the attraction of ants to the pheromone is very
intense, then a suboptimal pattern may emerge, as the ants crowd together in a sort of
pointless conformity. At the edge, though, at the very edge of chaos were the
parameters are tuned correctly, the ants will explore and follow the pheromone
signals, and wander from the swarm, and come back to it, and eventually coalesce into
a pattern that is, most of the time, the shortest, most efficient path from any given
point to any other.
In these real and simulated examples of insect accomplishments, we see optimization
of various types, whether clustering items [41,42] or finding the shortest path through
a landscape, with certain interesting characteristics. None of these instances include
global evaluation of the situation: an insect can only detect its immediate environment.
In contrast, optimization traditionally requires some method for evaluating the fitness
of a solution, which seems to require that candidate solutions be compared to some
standard, which may be a desired goal state (e.g. classical Genetic Algorithms) or the
fitness of other solutions (e.g. co-evolved Genetic Algorithms). The bottom-up
methods of the insect societies however, permit no evaluation – no ant knows how
well the swarm is doing. In general, the method of pheromone indirect communication
means that a more successful path will be somewhat more attractive, with an
autocatalytic accumulation of pheromone resulting in the population’s convergence on
the most-fit behaviour – all done at the local level [24].
If an ant colony on his cyclic way from the nest to a food source (and back again), has
only two possible branches around an obstacle, one bigger and the other smaller (the
bridge experiment [7,52]), pheromone will accumulate – as times passes – on the
shorter path, simple because any ant that sets out on that path will return sooner,
passing the same points more frequently, and via that way, reinforcing the signal of
that precise branch. Even if as we know, the pheromone evaporation rate is the same
in both branches, the longer branch will faster vanish his pheromone, since there is not
enough critical mass of individuals to keep it. On the other hand – in what appears to
be a vastly pedagogic trick of Mother Nature - evaporation plays a critical role on the
society. Without it, the final global decision or the phase transition will never happen.
Moreover, without it, the whole colony can never adapt if the environment suddenly
changes (e.g., the appearance of a third even shorter branch). While pheromone
reinforcement plays a role as system’s memory, evaporation allows the system to
adapt and dynamically decide, without any type of centralized or hierarchical control.
Similar social foraging capabilities have also been a source of inspiration to some
authors in the area of distributed Optimization and Control. Based on the biology and
physics underlying the foraging behaviour of E. coli bacteria, Passino and Liu [36,
28,29] exploit a variety of bacterial swarming and social foraging behaviours,
discussing how the control system on the E. coli dictates how foraging should
proceed. From here, an algorithmic model is presented that emulates the distributed
optimization process of the entire colony, with applications to a simple multipleextremum function minimization problem, as well as – in a brief final discussion – to
develop adaptive controllers and cooperative control strategies for autonomous
vehicles, a study later extended to noisy environments [29]. Other authors used


similar bacterial colony approaches for the optimization of Mobile Robot path
planning [46]. At a more theoretical level, Panait and Luke [35] presented recently a
pheromone-based model for collaborative foraging, and in a very interesting previous
work even studied how social foraging behaviours could computationally evolve [34].
Finally, other related paradigms and biology-inspired computing models [11] involve
the research and developing of Artificial Immune Systems (AIS) [15,10]. The
biological immune system is a highly parallel and distributed adaptive system. It uses
learning, memory, and associative retrieval to solve recognition and classification
tasks. In particular, it learns to recognize relevant patterns, remember patterns that
have been seen previously, and use combinations to construct pattern detectors
efficiently. These remarkable information-processing abilities of the immune system
provide important aspects in the field of computation, as in the specific case of agentbased AIS systems provided by Grilo, Caetano and Rosa [18].
All these above mentioned aspects show how vital can be the study of social foraging
for the development of new distributed search algorithms, and the construction of
social cognitive maps, with interesting properties in collective memory, collective
decision-making and swarm pattern detection. In the present work, we consider how
mechanisms inspired by social insects, in particular chemical signaling, may be used
to control large homogeneous populations of simple ant-like agents, so that the
collective result of their individual behaviors is the achievement of a particular global
goal. Specifically, the agents are required to move from a randomly distributed initial
spatial configuration to a tightly constrained configuration, as a 3D landscape peak. In
order to show that the system is highly robust as well extremely adaptive, we study the
colony behavior when one landscape suddenly changes into another. To achieve it, we
present a simple extended computer model based on pheromone distribution, which
represents the memory of the recent history of the swarm, and in a sense it contains
information which the individual ants are unable to hold or transmit. There is no direct
communication between the organisms but a type of indirect communication through
the pheromonal field. In this case, the environment plays a critical role on contextual
learning among the collective. On the other hand, these artificial ants sense and act
locally, so that control in the population as a whole is fully decentralized. Since, the
main aspects relate to self-organization and stigmergy, we first introduce them in some
detail as well as some universal guidelines that we consider relevant to the
development of highly distributed swarm systems of any kind.

2 Self-Organization and Stigmergy
Many structures built by social insects are the outcome of a process of selforganization, in which the repeated actions of the insects in the colony interact over
time with the changing physical environment to produce a characteristic end state
[20]. A major mediating factor is stigmergy [52], the elicitation of specific
environment-changing behaviors by the sensory effects of local environment changes
produced by previous and past behavior of the whole community. Stigmergy is a class
of mechanisms that mediate animal-animal interactions through artifacts or via indirect


communication, providing a kind of environmental synergy, information gathered
from work in progress, a distributed incremental learning and memory among the
society. In fact, the work surface is not only where the constituent units meet each
other and interact, as it is precisely where a dynamical cognitive map could be formed,
allowing for the embodiment of adaptive memory, cooperative learning and
perception [39]. Constituent units not only learn from the environment as they can
change it over time. Its introduction in 1959 by Pierre-Paul Grassé1 made it possible to
explain what had been until then considered paradoxical observations: In an insect
society individuals work as if they were alone while their collective activities appear
to be coordinated. The stimulation of the workers by the very performances they have
achieved is a significant one inducing accurate and adaptable response. The phrasing
of his introduction of the term is worth noting (translated to English in [20]):
The coordination of tasks and the regulation of constructions do not depend directly
on the workers, but on the constructions themselves. The worker does not direct his
work, but is guided by it. It is to this special form of stimulation that we give the name
Stigmergy (stigma - wound from a pointed object, and ergon - work, product of labor
= stimulating product of labor).
2.1 The role of the Environment
These self-organizing complex systems typically are comprised of a large number of
frequently similar components or events, occurring at a common environment. Trough
their process, a behavioral or physical pattern at the global-level of complexity,
emerges solely from numerous interactions. As provided by Bonabeau and colleagues
[7,9], self-organization (hereafter SO) refers to a set of dynamical mechanisms
whereby structures appear at the global level of a system from interactions among its
lower-level components. Moreover, the rules specifying the interactions among the
system’s constituent units are executed on the basis of purely local information,
without reference to the global pattern, which is an emergent property of the system
rather than a property imposed upon the system by an external ordering influence. On
the other hand, because the global (collective) properties of the system often defy
intuitive understanding of their origins, those properties may seem to appear
mysteriously. There is nothing mystical or unscientific about their emergence,
however [9]. Stigmergy is in fact, a well know example of self-organization in
biological systems, and can provide not only vital clues in order to understand how the
components can interact to produce a complex pattern, as can pinpoint simple
biological non-linear rules to achieve improved artificial intelligent systems, as those
found within Swarm Intelligence [42,8]. A paradigmatic case is provided by the
emergence of self-organization in social insects, by way of indirect interactions. An
example could be provided by two individuals, who interact indirectly when one of

1

Grassé, P.P.: La reconstruction du nid et les coordinations inter-individuelles chez
Bellicositermes natalensis et Cubitermes sp. La théorie de la stigmergie : Essai
d’interpretation des termites constructeurs. Insect Sociaux (1959), 6, 41-83.


A

t=1

t=10

t=30

t=40

t=50

t=80

t=100

t=120

t=150

t=200

t=500

t=1000

t=1500

t=2000

t=2500

t=3000

t=3500

t=4000

t=4500

t=5000

t=5500

t=6000

B

Fig.1 – One swarm (3000 ants) is thrown to
explore Kafka digital image (A) for 6000
iterations. At t=100, the Kafka image habitat is
replaced by Red Ant image (B). Evolutions of
swarm cognitive maps (pheromonal fields) are
shown for several iterations. Darker pixels
correspond to higher concentrations of
pheromone. Results from [39,40].

them modifies the environment and the other responds to the new environment at a
later time. Grassé showed that the coordination and regulation of building activities do
not depend on the workers themselves but are mainly achieved by the nest structure: a
stimulating configuration triggers the response of a termite worker, transforming the
configuration into another configuration that may trigger in turn another (possibly
different) action performed by the same termite or any other worker in the colony
(qualitative stigmergy). Another well-known illustration of how stigmergy and selforganization can be combined into more subtle adaptive behaviors is recruitment in
social insects. Self-organized trail laying via pheromone by individual ants
(quantitative stigmergy) is a way of modifying the environment to communicate with
nest mates that follow such trails [12,39]. It appears that task performance by some
workers decreases the need for more task performance: for instance, nest cleaning by
some workers reduces the need for nest cleaning [8]. Therefore, nest mates
communicate to other nest mates by modifying the environment (cleaning the nest),
and nest mates respond to the modified environment (by not engaging in nest
cleaning); that is, stigmergy.
Division of labor is another paradigmatic phenomenon of stigmergy. Simultaneous
task performance (parallelism) by specialized workers is believed to be more efficient
than sequential task performance by unspecialized workers [21]. Parallelism avoids
task switching, which costs energy and time. A key feature of division of labor is its
plasticity, being rarely rigid [43]. The ratios of workers performing the different tasks
that maintain the colony’s viability and reproductive success can vary in response to
internal perturbations or external challenges. In other words, local perception and


action taken by each agent can evolve adaptive and flexible problem-solving
mechanisms, or emerge communication among many parts, at the colony level.
Another crucial example is how ants form piles of items such as dead bodies
(corpses), larvae, or grains of sand. There again, stigmergy is at work: ants deposit
items at initially random locations. When other ants perceive deposited items, they are
stimulated to deposit items next to them, being this type of cemetery clustering
organization and brood sorting a type of self-organization and adaptive behavior.
Models of nest building in wasps [50,51] for instance, were already described, in
which wasp-like agents are stimulated to deposit bricks when they encounter specific
configurations of bricks: depositing a brick modifies the environment and hence the
stimulatory field of other agents. These asynchronous automata (designed by an
ensemble of algorithms) move in a 3D discrete space and behave locally in space and
time on a pure stimulus-response basis. There are other types of examples (e.g. prey
collectively transport), yet stigmergy is also present: ants change the perceived
environment of other ants (the colony cognitive map, according to [12,32,33]), and in
every example, the environment serves as medium of communication.
2.2 The role of Positive and Negative Feedbacks
Being these systems comprised of a large number of frequently similar components or
events, the principal challenge is to understand how the components interact to
produce a complex pattern [9]. Understanding the nature and properties of SelfOrganization (SO), and how it works, is not only crucial as can be of great insight. A
suitable approach is to first understand the two basic modes of interaction among the
components of SO systems: positive and negative feedback. Bonabeau et al [7,9]
identified four basic ingredients (negative and positive feedbacks, the amplification of
fluctuations, and the presence of multiple interactions) and three characteristic
signatures (the creation of spatiotemporal structures in an initially homogeneous
medium, the possible attainability of different stable states, i.e. multi-stability, and the
existence of parametrically determined bifurcations). Let’s now described the two
most important in some detail:
- Negative feedback, f-: a mechanism familiar to biologists, commonly used to
stabilize physiological processes (homeostasis) and avoid undesirable fluctuations. A
well-known example is the regulation of blood sugar levels, a process that proceeds
smoothly in most people but functions abnormally in diabetics. Blood sugar levels are
regulated by a negative feedback mechanism involving the release of insulin. An
increase in blood glucose following ingestion of sugary meal quickly triggers the
release of insulin from the pancreas, resulting in a number of physiological effects,
including the counteraction of the increase in blood-sugar level. In this case, we see facting to maintain the status quo by damping large fluctuations in blood glucose level.
A similar example involves the homeostatic regulation of body temperature. If the
organism experiences a thermal stress, then a discrepancy between the organism’s
actual body temperature (monitored by sensitive receptors in the hypothalamic area of
the brain) and his internal thermal set point, triggers various behavioral and


physiological responses. In both instances, the individual acquires and processes
information that elicits a negative feedback response, that is, a small perturbation
applied to the system triggers an opposing response that counteracts the perturbation.
This opposing response is normally set in action globally on the entire system. In the
first case an increase in blood glucose triggers a compensating response leading to a
decrease in blood glucose. In the second case, a decrease in body temperature results
in responses that increase body temperature. In other cases, negative feedback plays a
continuously role, without the necessity of having increase or decrease signals in
order to trigger it. A well-known example is provided by pheromone evaporation on
the self-organization of trails in some social insects, which acts globally and
continuously over time. Without evaporation the colony use of pheromone is useless,
since the whole system keeps reinforcing the same path (solution) over and over again,
being extremely fragile – and not robust - to new situations (e.g. landscape sudden
change). Other similar example is provided by people walking in footpaths, where
physical or weather conditions, such as erosion, play the role of negative feedback.
The overall final result however, is a dynamical robust memory [39,40] (a kind of
collective conscience) shared by a community seeking to find the “right”, appropriate
or aesthetical path, for instance, between two villages through a dense forest. What
strikes from this example, is that: 1) the community arrive at a common path, without
direct communication, 2) the signal reinforcement (see positive feedback) is made
discontinuously in time and, 3) at every new surprise (a landslide over parts of the
trail, for instance), rapidly the distributed whole of people locally finds a new solution
to re-connect the entire path again and again. As a mnemonically brain activation, a
sort of environmental synergy. The entire social cognitive map as the sum of these
distributed cues, is thus changing over time, and intelligently adapting, or selforganizing to new unforeseen environmental situations without a global master plan.
As if the environment is cooperating directly with the whole un-corporated mental
map, to form an entirely new, alive and embodied phenomenon. Without landslides or
obstacles like falling trees and people “getting lost” finding new solutions, the path
will never re-organize to explore new geographical areas, and without having people
locally to reinforce cues, the path in the first place will never emerge. For better or for
worse, without negative feedback the systems stables or freezes itself on a particular
configuration, an entire trail that never changes, keeps reinforcing itself and will never
evolve over time. Simultaneously and equally important, the SO system needs
components (people, in this case) that reinforce previous solutions (exploitation of the
search space) and components that more or less randomly, having or not a purpose to
it, subject themselves to areas “off the beaten track” (exploration), ideal to attract with their first and still fragile cues - the entirely community into a new global solution
in case of a sudden change or crisis (landslides). Often, components of the SO system
can do both, based on their non-linear response thresholds to environmental stimulus
[41], as in the model presented later.
- Positive feedback, f+: in contrast to negative feedback, positive feedback generally
promotes changes in the system (the majority of SO systems use them). The explosive
growth of the human population provides a familiar example of the effect of positive
feedback. The snowballing autocatalytic effect of f+ takes an initial change in a system


(due to amplification of fluctuations; a minimal and natural local cluster of objects
could be a starting point) and reinforces that change in the same direction as the initial
deviation. Self-enhancement, amplification, facilitation, and autocatalysis are all terms
used to describe positive feedback [9]. Another example could be provided by the
clustering or aggregation of individuals. Many birds, such as seagulls nest in large
colonies. Group nesting evidently provides individuals with certain benefits, such as
better detection of predators or greater ease in finding food. The mechanism in this
case is imitation2: birds preparing to nest are attracted to sites where other birds are
already nesting, while the behavioral rule could be synthesized as “I nest close where
you nest”. The key point is that aggregation of nesting birds at a particular site is not
purely a consequence of each bird being attracted to the site per se. Rather, the
aggregation evidently arises primarily because each bird is attracted to others (check
for further references on [7,9]). On social insect societies, f+ could be illustrated by
the pheromone reinforcement on trails, allowing the entire colony to exploit some past
and present solutions. Generally, as in the above cases, positive feedback is imposed
implicitly on the system and locally by each one of the constituent units. Fireflies
flashing in synchrony [49] follow the rule, “I signal when you signal”, fish traveling in
schools abide by the rule, “I go where you go”, and so forth. In humans, the
“infectious” quality of a yawn of laughter is a familiar example of positive feedback of
the form, “I do what you do”. Seeing a person yawning3, or even just thinking of
yawning, can trigger a yawn [9]. There is however one associated risk, generally if f+
acts alone without the presence of negative feedbacks, which per si can play a critical
role keeping under control this snowballing effect, providing inhibition to offset the
amplification and helping to shape it into a particular pattern. Indeed, the amplifying
nature of f+ means that it has the potential to produce destructive explosions or
implosions in any process where it plays a role. Thus the behavioral rule may be more
complicated than initially suggested, possessing both an autocatalytic as well as an
antagonistic aspect. In the case of fish [9], the minimal behavioral rule could be “I
nest where others nest, unless the area is overcrowded”. In this case both the positive
and negative feedback may be coded into the behavioral rules of the fish. Finally, in
other cases one finds that the inhibition arises automatically, often simply from
physical constraints.
Since in SO systems their organization arises entirely from multiple interactions, it is
of critical importance to question how organisms acquire and act upon information
[9]. Basically through two forms: a) information gathered from one’s neighbors, and
b) information gathered from work in progress, that is, stigmergy. In the case of
animal groups, these internal interactions typically involve information transfers
between individuals. Biologists have recently recognized that information can flow
2
3

See also on this subject the seminal sociological work of Gabriel Tarde; Tarde, G., Les Lois
de l’Imitation, Eds. du Seuil (2001), 1st Edition – Eds. Alcan, Paris, 1890.
Similarly, Milgram et al (Milgram, Bickerman and Berkowitz, "Note on the Drawing Power
of Crowds of Different Size", Journal of Personality and Social Psychology, 13, 1969)
found that if one person stood in a Manhattan street gazing at a sixth floor window, 20% of
pedestrians looked up; if five people stood gazing, then 80% of people looked up.


within groups via two distinct pathways – signals and cues. Signals are stimuli shaped
by natural selection specifically to convey information, whereas cues are stimuli that
convey information only incidentally [9]. The distinction between signals and cues is
illustrated by the difference ant and deer trails. The chemical trail deposited by ants as
they return from a desirable food source is a signal. Over evolutionary time such trails
have been molded by natural selection for the purpose of sharing with nestmates
information about the location of rich food sources. In contrast, the rutted trails made
by deer walking through the woods is a cue, not shaped by natural selection for
communication among deer but are a simple by-product of animals walking along the
same path. SO systems are based on both, but whereas signals tends to be
conspicuous, since natural selection has shaped signals to be strong and effective
displays, information transfer via cues is often more subtle and based on incidental
stimuli in an organism’s social environment [45].

3 A Swarm Cognitive Map Model for Dynamic Landscapes
As we shall see, the distribution of the pheromone represents the memory of the recent
history of the swarm, and in a sense it contains information which the individual ants
are unable to hold or transmit. There is no direct communication between the
organisms but a type of indirect communication through the pheromonal field. In fact,
ants are not allowed to have any memory and the individual’s spatial knowledge is
restricted to local information about the whole colony pheromone density. In order to
design this behaviour, one simple model was adopted [12], and extended due to
specific constraints of the present proposal, in order to deal with 3D dynamic
landscapes. As described by Chialvo and Millonas in [12], the state of an individual
ant can be expressed by its position r, and orientationθ. Since the response at a given
time is assumed to be independent of the previous history of the individual, it is
sufficient to specify a transition probability from one place and orientation (r,θ) to the
next (r*,θ*) an instant later. In previous works by Millonas [33,32], transition rules
were derived and generalized from noisy response functions, which in turn were found
to reproduce a number of experimental results with real ants. The response function
can effectively be translated into a two-parameter transition rule between the cells by
use of a pheromone weighting function (Eq.1):

σ
W (σ ) = 1 +
1 + γσ

β

(1)

This equation measures the relative probabilities of moving to a cite r (in our context,
to a cell) with pheromone density σ(r). The parameter β is associated with the
osmotropotaxic sensitivity, recognised by Wilson [54] as one of two fundamental
different types of ant’s sense-data processing. Osmotropotaxis, is related to a kind of
instantaneous pheromonal gradient following, while the other, klinotaxis, to a
sequential method (though only the former will be considered in the present work as in
[12]). Also it can be seen as a physiological inverse-noise parameter or gain. In
practical


F0a - 3D view

F0a - 2D view

F0b - 3D view

F0b - 2D view

F1 - 3D view

F1 - 2D view

F2 - 3D view

F2 - 2D view

F3 - 3D view

F3 - 2D view

F4 - 3D view

F4 - 2D view

Fig.2 – Three-dimensional views (3D) and
respective landscapes views (2D) of several
test functions used in our analysis [38]. White
pixels correspond to high peaks, while darker
ones represent deep valleys (F0-F4) or holes
(F6). Check table 1 in section 4.

F6 - 2D view

F6 - 2D view

t=0

t=0

t=500

t=500

t=50

t=50

t=1000

t=1000

t=100

t=100

Fig.3 – maxF0a. Pheromone distribution
(Social Cognitive Maps) for t=0, 50, 100, 500
and 1000 time steps, of 3000 ants exploring
function F0a on a 100 x 100 toroidal grid (1st
and 3rd column: darker pixels correspond to
higher concentrations). Columns 2 and 4
correspond to the geographical place where
agents are situated (each black pixel is an
ant). At t=100, the highest peak is already
surrounded by agents while convergence
proceeds. Processing time equals to 54 s.


Pik =

W (σ i )w(∆ i )
W (σ j )w(∆ j )
j/k

(2)

terms, this parameter controls the degree of randomness with which each ant follows
the gradient of pheromone. On the other hand, 1/ is the sensory capacity, which
describes the fact that each ant’s ability to sense pheromone decreases somewhat at
high concentrations. In addition to the former equation, there is a weighting factor
w(∆θ), where ∆θ is the change in direction at each time step, i.e. measures the
magnitude of the difference in orientation. As an additional condition, each individual
leaves a constant amount η of pheromone at the cell in which it is located at every
time step t. This pheromone decays at each time step at a rate k. Then, the normalised
transition probabilities on the lattice to go from cell k to cell i are given by Pik (Eq. 2,
[12]), where the notation j/k indicates the sum over all the surrounding cells j which
are in the local neighbourhood of k. ∆i measures the magnitude of the difference in
orientation for the previous direction at time t-1. That is, since we use a
neighbourhood composed of the cell and its eight neighbours, ∆i can take the discrete
values 0 through 4, and it is sufficient to assign a value wi for each of these changes of
direction. Chialvo et al. used the weights of w0 =1 (same direction), w1 =1/2, w2 =1/4,
w3 =1/12 and w4 =1/20 (U-turn). In addition, coherent results were found for η=0.07
(pheromone deposition rate), k=0.015 (pheromone evaporation rate), β=3.5
(osmotropotaxic sensitivity) and γ =0.2 (inverse of sensory capacity), where the
emergence of well defined networks of trails were possible. Except when indicated,
these values will remain in the following framework. As an additional condition, each
individual leaves a constant amount η of pheromone at the cell in which it is located at
every time step t. Simultaneously, the pheromone evaporates at rate k, i.e., the
pheromonal field will contain information about past movements of the organisms, but
not arbitrarily in the past, since the field forgets its distant history due to evaporation
in a time τ ≅ 1/k. As in past works, toroidal boundary conditions are imposed on the
lattice to remove, as far as possible any boundary effects (e.g. one ant going out of the
grid at the south-west corner, will probably come in at the north-east corner).
In order to achieve emergent and autocatalytic mass behaviours around specific
locations (e.g., peaks or valleys) on the habitat, instead of a constant pheromone
deposition rate η used in [12], a term not constant is included. This upgrade can
significantly change the expected ant colony cognitive map (pheromonal field). The
strategy follows an idea implemented earlier by Ramos [39] (fig. 1), while extending
the Chialvo model into digital image habitats, aiming to achieve a collective
perception of those images by the end product of swarm interactions. The main
differences to the Chialvo work, is that ants now move on a 3D discrete grid,
representing the functions which we aim to study (fig. 2) instead of a 2D habitat, and
the pheromone update takes in account not only the local pheromone distribution as
well as some characteristics of the cells around one ant. In here, this additional term
should naturally be related with specific characteristics of cells around one ant, like
their altitude (z value or function value at coordinates x,y), due to our present aim. So,
our pheromone


Algorithm.
High-level description of the SWARM SEARCH
algorithm proposed
t=0

t=0

t = 500

t = 500

t = 1000

t = 1000

t = 1010

t = 1010

t = 1050

t = 1050

t = 1080

t = 1080

t = 1100

t = 1100

/* Initialization */
For all agents do
Place agent at randomly selected site
End For
/* Main loop */
For t = 1 to tmax do
For all agents do
/* According to Eqs. 1 and 2 (section 3) */
Compute W( ) and Pik
Move to a selected neighboring site not
occupied by other agent
/* According to Eq. 3 (section 3) */
Increase pheromone at site r:
Pr= Pr+[η+p( [r]/ max)]
End For
Evaporate pheromone by K, at all grid sites
End For
Print location of agents
Print pheromone distribution at all sites
/* Values of parameters used in experiments */
k = 0.015, η = 0.07, β =3.5, =0.2,
p = 1.9, tmax = 500, 600, 1000 or 1150 steps.
/* Useful references */
Check [39], [41], [12], [33] and [32].

Fig.4 – maxF0a => maxF0b. Social
evolution from maximizing function F0a to
maximizing function F0b. In the first 1000
time steps the ant colony explores function
F0a, while suddenly at t=1001, function F0b
is used as the new habitat. Pheromone
distribution (Social Cognitive Maps) for t =
0, 500, 1000, 1010, 1050, 1080, 1100 and
1150 time steps, of 3000 ants exploring
function F0a and F0b on a 100 x 100
toroidal grid are shown. Already at t=1010,
the old highest peak on the right suffers a
radical erosion, on the presence of ants (they
start to explore new regions). As time passes
the majority of the colony moves to the new
peak, on the left. The strategy pseudo-code is
given above.


t = 50

t = 50

t = 150

t = 150

t = 250

t = 250

t = 300

t = 300

t = 350

t = 350

t = 400

t = 400

t = 450

t = 450

t = 500

t = 500

Fig.5 – maxF0a => minF0a. Maximizing function F0a during 250 time steps and then
minimizing it for t ≥ 251. Pheromone distribution (Social Cognitive Maps) for t = 50, 150,
250, 300, 350, 400, 450 and 500 time steps, of 2000 ants exploring function F0a on a 100 x
100 toroidal grid are shown. Already at t=300, the highest peak on the right suffers a radical
erosion, on the presence of ants (they start to explore new regions). As time passes the
majority of the colony moves to the deep valley, on the left. Parameters are different from
those used in Figs. 3-4 (check table 2).

T =η + p

∆[i ]
∆ max

(3)

deposition rate T, for a specific ant, at one specific cell i (at time t), should change to a
dynamic value (p is a constant = 1.93) expressed by equation 3. In this equation, max
= | zmax – zmin |, being zmax the maximum altitude found by the colony so far on the
function habitat, and zmin the lowest altitude. The other term [i] is equivalent to (if
our aim is to minimize any given landscape): [i] = | zi – zmax |, being zi the current
altitude of one ant at cell i. If on the contrary, our aim is to maximize any given
landscape, then we should instead use [i] = | zi – zmin |. Finally, please notice that if
our landscape is completely flat, results expected by this extended model will be equal
to those found by Chialvo and Millonas in [12], since [i]/∆max equals to zero. In this
case, this is equivalent to say that only the swarm pheromonal field is affecting each


t = 20

t = 20

t = 100

t = 100

t = 300

t = 300

t = 320

t = 320

t = 400

t = 400

t = 500

t = 500

t = 600

t = 600

Fig.6 – minF6 => maxF0a. Minimizing
function F6 during 300 time steps and then
maximizing function F0a for t ≥ 301.
Pheromone distribution (Social Cognitive
Maps) for t = 20, 100, 300, 320, 400, 500,
and 600 time steps, of 3000 ants exploring
function F6 and F0a on a 100 x 100 toroidal
grid are shown. Parameters are different from
those used in Figs. 3-4 (check table 2).

ant choices, and not the environment - i.e. the expected network of trails depends
largely on the initial random position of the colony, and in trail clusters formed in the
initial configurations of pheromone. On the other hand, if this environmental term is
added a stable and emergent configuration will appear which is largely independent on
the initial conditions of the colony and becomes more and more dependent on the
nature of the current studied landscape itself. As specified earlier in section 2.1, the
environment plays an active role, in conjunction with continuous positive and negative
feedbacks (section 2.2) provided by the colony and their pheromone, in order to
achieve a stable emergent pattern, memory and distributed learning by the community.

4 Experimental Setup and Results
In order to test the dynamical behaviour of this new Swarm Search algorithm
presented earlier in section 3 (pseudo-code in fig. 4), we have used classical test
functions (table 1) drawn from the literature in Genetic Algorithms, Evolutionary
strategies and global optimization [38], several of them graphically accessible in fig.
2. Function F0a represents one deep valley and one peak, while F0b his opposite.


Table 1. Classical test functions used in our analysis from MATLAB reference manual [38]

Function ID
F0a

Equation

f 0 a ( x ) = x1.e

F0b

F3

F5
F6

f 1b ( x ) =
f 2 (x ) =

n −1
i =1

i =1

i =1
n

f1a ( x ) =

i =1

xi2

xi2
i.xi2

n

i

i =1

j =1

(

x i2

n

− 0.2

n

f1 ( x ) =

F2

F4

i =1

f 0b (x ) = − x1.e

F1

n

− 0.2

2

xj

)

100. xi+1 − xi2 + (1 − xi )

f 6 ( x ) = 10.n +
f 7 (x ) =

n
i =1
n
i =1

2

(x

2
i

2

)

− 10. cos(2.π .xi )

(

− xi .sin | xi |

)

Function F1 represents De Jong’s function 1 and his one of the simplest. It is
continuous, convex and unimodal; xi is in the interval [-5.12; 5.12] and the global
minimum is at xi=0. Function F2 represents an axis parallel hyper-ellipsoid similar to
De Jong’s function 1. It is also know as the weighted sphere model. Again it is
continuous, convex and unimodal in the interval xi
[-5.12; 5.12], with global
minimum at xi=0. Function F3 represents an extension of the axis parallel hyperellipsoid (F2), also know as Schwefel’s function 1.2. With respect to the coordinate
axes this function produces rotated hyper-ellipsoids; xi is in the interval [-65.536;
65.536] and the global minimum is at xi=0. Likewise F2, it is continuous, convex and
unimodal. Function F4 represents the well-know Rosenbrock’s valley or De Jong’s
function 2. Rosenbrock’s valley is a classic optimization problem. The global
optimum is inside a long, narrow, parabolic shaped flat valley. To find the valley is
trivial, however convergence to the global optimum is difficult and hence this problem
has been repeatedly used in assess the performance of optimization algorithms; xi is in
the interval [-2.048; 2.048] and the global minimum is at xi=0. Function F5 represents
the Rastrigin’s function 6. This function is based on De Jong’s function 1 with the
addition of cosine modulation to produce many local minima. Thus, the test function
is highly multimodal. However, the location of the minima are regularly distributed.
As in F1, xi is in the interval [-5.12; 5.12] and the global minimum is at xi=0. Finally,


Table 2. Parameters used for different test sets (check section 3 and figures 3-6 for details)

Test
maxF0a
maxF0a =>
maxF0b
maxF0a =>
minF0a
minF6 =>
maxF0a

η

β

Fig.
3
4

N ants
3000
3000

tmax
1000
1150

k
0.015
0.015

0.07
0.07

3.5
3.5

0.2
0.2

p
1.93
1.93

5

2000

500

1.000

0.10

3.5

0.2

1.90

6

3000

600

1.000

0.01

3.5

0.2

1.90

F6 represents Schwefel’s function 7, being deceptive in that the global minimum is
geometrically distant, over the parameter space, from the next best local minima.
Therefore, the search algorithms are potentially prone to convergence in the wrong
direction; xi is in the interval [-500; 500] and the global minimum is at xi=420,9687
while f(x)=n.418,9829. In our tests, n=2. Within this specific framework we have
produced several run tests using different test functions, some of which are presented
here trough figures 3 to 6. The parameters used are shown on table 2. The simplest
test was the first one (fig.3) where we forced the colony to search for the maximal
peak in function F0a, during 1000 time steps. The other tests were harder, since they
include not only different purposes simultaneously (maximizing and minimizing) as
well as different landscapes that changed dynamically on intermediate swarm search
stages (e.g., fig. 6).
4.1 SWARM SEARCH versus BACTERIAL FORAGING algorithms
In order to further analyze the collective behavior of the present proposal, we make a
comparison between the ant-like Swarm Search Algorithm (SSA) and the Bacterial
Foraging Optimization Algorithm (BFOA), on the dominion of function optimization.
BFOA was selected since it represents an earlier proposal for function optimization as
well based on natural foraging capacities. Presented by Passino at IEEE Control
Systems Magazine in 2002 [36] and later that year in the Journal of Optimization
Theory and Applications [28], the author for the purpose of a simple but powerful
illustrative example, used his algorithm to find the minimum of two complex functions
Jcc, described in [36], page 60. Further material, as the MATLAB code of his
algorithm and the tri-dimensional functions experimented, can also be found on the
web address of a recent book from the same author (Biomimicry for Optimization,
Control and Automation, Springer-Verlag, London, UK, 2005), at
http://www.ece.osu.edu/ ~passino/ICbook/ ic_index.html. Passino uses S=50 bacteriabased agents, during four generations. In each generation, and has a requirement of his
algorithm, each agent enters a chemotaxis loop (see page 61 [36]), performing Nc=100
chemotactic (foraging) steps. Thus the algorithm – for the precise application – runs
for t=400 time steps, which make us believe that a fair comparison can be make in
regard of the parameter values we use. The two functions represent what Passino
designates by nutrient concentration landscapes (see fig. 7, first row – the web
address also contains his MATLAB code used in the two functions, where


Passino F1 – 3D

Passino F1 – 2D

Passino F2 – 3D

Passino F2 – 2D

t = 100

t = 200

t = 300

t = 400

t = 100

t = 200

t = 300

t = 400

t = 100

t = 200

t = 300

t = 400

t = 100

t = 200

t = 300

t = 400

t = 100

t = 200

t = 300

t = 400

Fig.7 – In the first row the test functions used by Passino [36,28]. In the second and third
rows, BFOA minimizing results respectively for F1 and F2. The graphics show the bacterial
motion trajectories (using 50 bacteria-like agents). In the fourth and fifth rows, SWARMSEARCH algorithm (SSA) minimizing results respectively for F1 and F2, and for the same
foraging time period. The graphics shows the pheromone distribution. In the last row, SSA is
requested to deal with two contradictory goals, i.e. to minimize F1 and then to maximize it. In
all these tests, SSA has used 50 ant-like agents. Check main text for the parameters used.
Habitat size equals 2 x [0,30].


Nutrientsfunc.m and Nutrientsfunc1.m are represented by different weights). His
function F2 (Nutrientsfunc1.m) has a zero value at [15,15] and decreases to
successively more negative values as you move away from that point, reaching a
plateau with the same value. Moreover, and for the purpose of discrete function
optimization, Passino [36] represented both functions by a discrete lattice (as well as
us in our past tests) with a size of 30 x 30 cells over the optimization domain (each
cell has a correspondent z or Jcc value). For these reasons and in order to keep a
coherent comparison, we shall use 50 ant-like agents in our SSA, on a 30 x 30 tridimensional habitat, for t=400 time steps, on both functions. We then run 3 tests. The
first is requested to minimize Passino’s function F1. The second test is requested to
minimize Passino’s function 2. Finally, and in order to prove the adaptive features of
our model, we requested SSA to deal with two contradictory goals, i.e. to minimize F1
and then to maximize it, over the same period of 400 time steps. As visible, SSA
quickly adapts to the different purposes. Over function F1, the pheromone
concentration is already intensely allocated at the right point at t=100 (and not in other
areas), while BFOA still explores different regions on the optimization domain. Over
function F2, the swarm quickly separates in different foraging groups, since there are
a large number of points with the minimal value. Finally over function F1 again, in the
final test (last row – fig. 7), SSA is able to process two different demands
(maximization followed by minimization) over the same foraging time period that
BFOA uses for F1 minimization. The parameters used in our experiments follows:
Nants=50, tmax=400, k =1 (pheromone evaporation rate), η=0.1 (pheromone deposition
rate), β=7 (this parameter controls how ants follow the pheromone gradient), =0.2,
and p=1.9. In test 1, β=6.

5 Conclusions
Evolution of mass behaviours on time are difficult to predict, since the global
behaviour is the result of many part relations operating in their own local
neighbourhood. The emergence of network trails in ant colonies, for instance, are the
product of several simple and local interactions that can evolve to complex patterns,
which in some sense translate a meta-behaviour of that swarm. Moreover, the
translation of one kind of low-level (present in a large number) to one meta-level is
minimal. Although that behaviour is specified (and somehow constrained), there is
minimal specification of the mechanism required to generate that behaviour; global
behaviour evolves from the many relations of multiple simple behaviours, without
global coordination (i.e. from local interactions to global complexity. One
paradigmatic and abstract example is the notion, within a specified population, of
common-sense, being the meta-result a type of collective-conscience. There is some
evidence that our brain as well as many other complex systems, operates in the same
way, and as a consequence collective perception capabilities could be derived from
emergent properties, which cannot be neglected in any pattern search algorithm. These
systems show in general, interesting and desirable features as flexibility (e.g. the brain
is able to cope with incorrect, ambiguous or distorted information, or even to deal


with unforeseen or new situations without showing abrupt performance breakdown) or
versability, robustness (keep functioning even when some parts are locally
damaged), and they operate in a massively parallel fashion. Present results point to
that type of interesting features. Although the current model is far from being
consistent with real ones, since only some type of mechanisms are considered, swarm
pheromonal fields reflect some convergence towards the identification of a common
goal in a purely decentralized from. Moreover, the present model shows important
adaptive capabilities, as in the presence of sudden changes in the habitat - our test
landscapes (fig. 3). Even if the model is able to quickly adapt to one specific
environment, evolving from one empty pheromonal field, habitat transitions point
that, the whole system is able to have some memory from past environments (i.e.
convergence is more difficult after learning and perceiving one past habitat). This
emerged feature of résistance, is somewhat present in many of the natural phenomena
that we find today in our society. In a certain sense, the distribution of pheromone
represents the collective solutions found so far (memory, risk avoidance, exploitation
behavior), while evaporation enables the system to adapt (tricks a decision,
explorative behavior), not only as in normal situations (a complex search landscape),
as well as when the landscape suddenly changes, moving the colony’s new target to a
new unexplored region. And again as noted by Langton [27,26], as in many complex
systems, only at the right intermediary regime, in here between contradictory
behaviors of exploration and exploitation, the swarm is able to quickly converge.
The recognizable results indicate that the collective intelligence is able to cope and
quickly adapt to unforeseen situations even when over the same cooperative foraging
period, the community is requested to deal with two different and contradictory
purposes. All these above mentioned aspects show how vital can be the study of social
foraging for the development of new distributed search algorithms, and the
construction of social cognitive maps, with interesting properties in collective
memory, collective decision-making and swarm pattern detection.
But the work could have important consequences in other areas. Perhaps, one of the
most valuable relations to explore is that of social foraging and evolution. For two
reasons; First, as described by Passino [36], natural selection tends to eliminate
animals with poor “foraging strategies” (methods for locating, handling, and ingesting
food) and favor the propagation of genes of those animals that have successful
foraging strategies since they are more likely to enjoy reproductive success (they
obtain enough food to enable them to reproduce). Logically, such evolutionary
principles have led scientists in the field of foraging theory to hypothesize that it is
appropriate to model the activity of foraging as an optimization process: A foraging
animal takes actions to maximize the energy obtained per unit time spent foraging, in
the face of constraints presented by its own physiology and by the environment.
Second, because there is an increasing recognition that natural selection and selforganization work hand in hand to form evolution, as defended by Kauffmann [22,23].
For example, anthropologist Jeffrey McKee [31,24] has described the evolution of
human brain as a self-organizing process. He uses the term autocatalysis to describe
how the design of an organism’s features at one point in time affects or even
determines the kinds of designs it can change into later. For example the angle of the
skull on the top of the spine left some extra space for the brain to expand. Thus the


evolution of the organism is determined not only by selection pressures but by
constraints and opportunities offered by the structures that have evolved so far. Also,
and back again in what regards the evolution of collectives, it is known that during the
evolution of life, there have been several transitions in which individuals began to
cooperate, forming higher levels of organization and sometimes losing their
independent reproductive identity (insect societies are one example). Several factors
that confer evolutionary advantages on higher levels of organization have been
proposed, such as Division of Labor and Increased Size. But recently, a new third
factor was added: Information Sharing [25]. Lachmann et al, illustrate with a simple
model how information sharing can result in individuals that both receive more
information about their environment and pay less for it.
Being social foraging essentially a self-organized phenomenon, the study of
computational foraging embedded with GA (Genetic Algorithm) like natural selection
can much probably enhance our understanding on the detailed forms of the
hypothetical equation: Evolution = Natural Selection + Self-Organization, and in the
precise role of each “variable”. As an example, current work by present authors,
include the research of variable population size swarms, as used similarly in
Evolutionary Computation (Fernandes and Rosa [17]), where each individual can have
a probability of making a child, as well to die, depending on his accumulated versus
spent energetic resources. The system as a whole, then proceeds on the search space
as a kind of distributed evolutionary swarm.

References
1. Bak, P., How Nature Works – The Science of Self-Organized Criticality, Springer-Verlag,
1996.
2. Ball, P., “Do Plants act like Computers?”, Nature News Service, MacMillan Magazines Ltd,
21 Jan. 2004.
3. Bassler, B.L, “Small Talk: Cell-to-Cell Communication in Bacteria”, Cell, Vol. 109, pp.
421-424, May 2002.
4. Ben-Jacob, E., Shochet, O., Tenenbaum, A., Cohen, I., Czirók, A., Vicsek, T., “Generic
Modelling of Cooperative Growth in Bacterial Colonies”, Nature, 368, pp. 46-49, 1994.
5. Ben-Jacob, E., “Bacterial Wisdom, Gödel’s Theorem and Creative Genomic Webs”, Physica
A, 248, pp. 57-76, 1998.
6. Ben-Jacob, E., Becker, I., Shapira, Y., Levine, H., “Bacterial Linguistic Communication and
Social Intelligence”, Trends in Microbiology, Vol. 12/8, pp. 366-372, 2004.
7. Bonabeau, E., Theralauz, G., Deneubourg, J.-L., Camazine S., “Self-Organization in Social
Insects”, Trends in Ecology and Evolution, 12(5), 188-193, 1997.
8. Bonabeau, E., Dorigo, M., Theraulaz, G., Swarm Intelligence: From Natural to Artificial
Systems, Santa Fe Institute in the Sciences of Complexity, Oxford Univ. Press, New York,
Oxford, 1999.
9. Camazine, S., Deneubourg, J.-L., Franks, N.R., Sneyd, J., Theraulaz, G., Bonabeau, E., SelfOrganization in Biological Systems, Princeton Studies in Complexity, Princeton University
Press, Princeton and Oxford, 2001.
10. Castro, L.N., Timmis, J.I., Artificial Immune Systems: A New Computational Intelligence
Approach, Springer-Verlag, 2002.
11. Castro, L.N., Von Zuben, F.J., Recent Developments in Biologically Inspired Computing,
Eds., Idea Group Publishing Inc., 2005.


12. Chialvo, D.R., Millonas, M.M., “How Swarms build Cognitive Maps”, In Steels, L. (Ed.):
The Biology and Technology of Intelligent Autonomous Agents, 144, NATO ASI Series,
439-450, 1995.
13. Chialvo, D.R., “Critical Brain Networks”, Physica A, 340, 4, 756, 2004.
14. Chowdhury, D., Nishinari, K., Schadschneider, A., “Self-Organized Patterns and Traffic
Flow in Colonies of Organisms: from Bacteria and Social Insects to Vertebrates”, special
issue on Pattern Formation, in Phase Transitions, Taylor and Francis, vol. 77, 601, 2004.
15. Dasgupta, D., Artificial Immune Systems and Their Applications, Ed., Springer-Verlag,
1999.
16. Farmer, J.D., “A Rosetta Stone for Connectionism”, in S. Forrest (Ed.), Emergent
Computation, Cambridge, MA: The MIT Press, 1991.
17. Fernandes, C., Rosa, A.C., “Study on Non-random Mating and Varying Population Size in
Genetic Algorithms using a Royal Road Function”, IEEE CEC´01, Proc. of the 2001 IEEE
Congress on Evolutionary Computation, pp. 60-66, 2001.
18. Grilo, A., Caetano, A., Rosa, A.C., “Agent based Artificial Immune System”, Proc.
GECCO-01, Vol. LBP, pp. 145-151, 2001.
19. Hofstadter, D.R., Gödel, Escher, Bach: An Eternal Golden Braid, New York: Basic Books,
1979.
20. Holland, O., Melhuish, C.: Stigmergy, “Self-Organization and Sorting in Collective
Robotics”, Artificial Life, Vol. 5, n. 2, MIT Press, 173-202, 1999.
21. Jeanne, R.L., “The Evolution of the Organization of Work in Social Insects”, Monit. Zool.
Ital., 20, 119.133, 1986.
22. Kauffmann, S.A., The Origins of Order: Self-Organization and Selection in Evolution,
New York: Oxford University Press, 1993.
23. Kauffmann, S.A., At Home in the Universe: The Search for the Laws of Self-Organization
and Complexity, New York: Oxford University Press, 1995.
24. Kennedy, J. Eberhart, Russel C. and Shi, Y., Swarm Intelligence, Academic Press, Morgan
Kaufmann Publ., San Diego, London, 2001.
25. Lachmann, M., Sella, G., Jablonka, E., “On Information Sharing and the Evolution of
Collectives”, Proc. of the Royal Society: Biological Sciences, 267, pp. 1265-1374, 2000.
26. Langton, C.G., “Computation at the Edge of Chaos”, Physica D, 42, pp. 12-37, 1990.
27. Langton, C.G., “Life at the Edge of Chaos”, in C.G. Langton (Ed.), Artificial Life II, Santa
Fe Institute Studies in the Sciences of Complexity, Vol. 10, Reading, MA: Addison-Wesely,
pp. 41-91, 1991.
28. Liu, Y., Passino, K.M., “Biomimicry of Social Foraging Bacteria for Distributed
Optimization: Models, Principles, and Emergent Behaviors”, Journal of Optimization
Theory and Applications, Vol. 115, nº3, pp. 603-628, Dec. 2002.
29. Liu, Y., Passino, K.M., “Stable Social Foraging Swarms in a Noisy Environment,” IEEE
Trans. on Automatic Control, Vol. 49, No. 1, pp. 30-44, Jan. 2004.
30. Maree, A.F.M., Hogeweg, P., “How Amoeboids Self-Organize into a Fruiting Body:
Multicellullar Coordination in Dictyostelium discoideum”, PNAS, vol. 98, nº 7, pp. 38793883, 2001.
31. McKee, J.K., The Riddled Chain: Change, Coincidence, and Chaos in Human Evolution,
Piscataway, NJ: Rutjers University Press, 2000.
32. Millonas, M.M., “A Connectionist-type model of Self-Organized Foraging and Emergent
Behavior in Ant Swarms”, J. Theor. Biol., nº 159, 529, 1992.
33. Millonas, M.M., “Swarms, Phase Transitions and Collective Intelligence”, In Langton,
C.G. (Ed.): Artificial Life III, Santa Fe Institute, Studies in the Sciences of Complexity, Vol.
XVII, Addison-Wesley, Reading, Massachusetts, 417-445, 1994.
34. Panait, L., Luke, S., “Evolving Foraging Behaviors”, Proc. of the Ninth Int. Conf. on the
Simulation and Synthesis of Living Systems, ALIFE-9, 2003.


35. Panait, L., Luke, S., “A Pheromone-based Utility Model for Collaborative Foraging”, Proc.
of the 2004 Conference on Autonomous Agents and Multiagent Systems, 2004.
36. Passino, K.M., “Biomimicry of Bacterial Foraging for Distributed Optimization and
Control”, IEEE Control Systems Magazine, pp. 52-67, June 2002.
37. Peak, D.A., West, J.D., Messinger, S.M., Mott, K.A., “Evidence for Complex, Collective
Dynamics and Emergent, Distributed Computation in Plants”, PNAS, Proc. of the National
Academy of Sciences, USA, 101, pp. 918-922, 2004.
38. Pohlheim, H, “Genetic Algorithm MATLAB Toolbox Test Functions”, MATLAB reference
manual, version 1.2, Mathworks, 1997.
39. Ramos, V., Almeida, F., “Artificial Ant Colonies in Digital Image Habitats: A Mass
Behavior Effect Study on Pattern Recognition”, In Dorigo, M., Middendorf, M., Stuzle, T.
(Eds.): From Ant Colonies to Artificial Ants - 2nd Int. Wkshp on Ant Algorithms, 113-116,
2000.
40. Ramos, V., “On the Implicit and on the Artificial - Morphogenesis and Emergent
Aesthetics in Autonomous Collective Systems”, in ARCHITOPIA Book, Art, Architecture
and Science, Institut D’Art Contemporain, J.L. Maubant et al. (Eds.), pp. 25-57, Chapter 2,
ISBN 2905985631 – EAN 9782905985637, France, Feb. 2002.
41. Ramos,V., Merelo, Juan J., “Self-Organized Stigmergic Document Maps: Environment as a
Mechanism for Context Learning”, in AEB’2002 – 1st Spanish Conf. on Evolutionary and
Bio-Inspired Algorithms, E. Alba, F. Herrera, J.J. Merelo et al. (Eds.), pp. 284-293, Centro
Univ. de Mérida, Mérida, Spain, 6-8 Feb. 2002.
42. Ramos, V., Abraham, A., “Evolving a Stigmergic Self-Organized Data-Mining”, in ISDA04, 4th Int. Conf. on Intelligent Systems, Design and Applications, Budapest, Hungary,
ISBN 963-7154-30-2, pp. 725-730, August 26-28, 2004.
43. Robinson, G.E., “Regulation of Vision of Labor in Insect Societies”, Annu. Rev. Entomol.,
37, 637-665, 1992.
44. Savill, N.J., Hogeweg, P., “Modelling Morphogenesis: From Single Cells to Crawling
Slugs”, J. Theor. Biology, 184, pp. 229-235, 1997.
45. Seeley, T.D., “The Honey Bee Colony as a Super-Organism”, American Scientist, 77, pp.
546-553, 1989.
46. Sierakowski, C.A., Coelho, L.S., “Path Planning Optimization for Mobile Robots based on
Bacteria Colony Approach” (in Portuguese), VIII Simpósio Brasileiro de Redes Neurais,
2004.
47. Silberman, S., “The Bacteria Whisperer”, Wired, 11-04, pp. 104-108, April 2003.
48. Sporns, O., Chialvo, D.R., Kaiser, M., Hilgetag, C.C., “Organization, Development and
Function of Complex Brain Networks”, Trends in Cognitive Sciences, Vol. 8, nº9, Sept.
2004.
49. Strogatz, S., Sync – The Emerging Science of Spontaneous Order, Theia, New York, 2003.
50. Theraulaz, G., Bonabeau, E., “Coordination in Distributed Building”, Science, 269, 686688, 1995.
51. Theraulaz, G., Bonabeau, E., “Modelling the Collective Building of Complex Architectures
in Social Sciences wit Lattice Swarms”, J. Theor. Biol., 1777, 381-400, 1995.
52. Theraulaz, G., Bonabeau, E., “A Brief History of Stigmergy”, Artificial Life, Vol. 5, n. 2,
MIT Press, 97-116, 1999.
53. Williams, H., Spatial Organisation of a Homogeneous Agent Population using Diffusive
Signalling and Role Differentiation, MSc Evolutionary & Adaptive Systems Thesis, COGS,
Univ. of Sussex, UK, Sept. 2002.
54. Wilson, E.O., The Insect Societies, Cambridge, MA., Belknap Press, 1971.
55. Wolfram, S., “Universality and Complexity in Cellular Automata”, Physica D, 10, pp. 1-35,
1984.


Ramos, v; fernandes, c; rosa, a. (2007): social cognitive maps, swarm collective perception and distributed search on dynamic landscapes

More Related Content

What's hot (11)

Viewers also liked (7)

Similar to Ramos, v; fernandes, c; rosa, a. (2007): social cognitive maps, swarm collective perception and distributed search on dynamic landscapes (20)

More from ArchiLab 7 (20)

Recently uploaded (20)

Ramos, v; fernandes, c; rosa, a. (2007): social cognitive maps, swarm collective perception and distributed search on dynamic landscapes