Report on Knowledge Modeling in Various applications in Traffic Systems

2014
Faculty of Computers and
Information
Department of Computer
Sciences
Cairo University
Knowledge Modeling
in Various applications
in Traffic Systems
Report Submitted by: Yomna Mahmoud Ibrahim Hassan
Report Submitted to : Prof. Dr. Hesham Hassan

2 sn tuoaiogf aou nig ngide lKgAssu edw nKg ngledMM egdeKwonK
Table of Contents
Introduction page 3
Knowledge Modeling Techniques page 3
Generic Tasks page 3
CommonKADS page 5
Traffic Systems page 7
CommonKADS and Traffic Systems page 7
MAS- CommonKADS page 11
Dynamic Virtual Organization page 13
Conclusions page 13
References page 14

Knowledge Modeling in Various
Applications in Traffic Systems
Introduction
This report entails a detailed study of the knowledge modeling techniques utilized in
previous research in the traffic domain. The finalized model reached within this report
was a combination of research presented in multiple papers with the goal of reaching
a comprehensive technique for knowledge modeling traffic systems. The techniques
used were enhanced over the year in order to incorporate changes within current
traffic systems. This report is divided into the following sections, we first start with an
introduction where we give a general idea about knowledge modeling within traffic
systems and why do we need to model such a system within a knowledge base.
Within the second section, we give an introduction about traffic systems components.
Afterwards, each section will include part of the knowledge model combined from
research papers. Conclusions are finally described in the last section.
Section 1: Knowledge Modeling Techniques
Knowledge engineering (KE) is considered as the discipline of identifying a structure
that can be re-utilized for any knowledge based system (KBS). The basic structure of
a KBS is described in Figure 1. A systematic design for the knowledge structure of
the system was to be achieved, later on called knowledge modeling. Two of the most
prominent knowledge modeling techniques are Generic Tasks (GT) and
CommonKADS.
Figure 1: Knowledge Based system structure
a- Generic Tasks
Generic tasks, created in the late 80s [1], are templates of problem-solving activities
that can be configured together to describe any intelligent activity. At least five
different types of knowledge can be identified within generic tasks. These types are:

- Tasks
- Problem-solving method
- Inferences
- Ontologies
- Domain Knowledge
These different knowledge model components are related as shown in Figure 2. Tasks
within GT are each domain and problem independent. Each Generic Task is a
specialized type of problem solving which is described functionally. By describing a
task functionally, a method for solving that task is fairly easy to construct. The
method then dictates the types of knowledge required to solve that problem.
Once you have decided to that a certain GT is fit to solve your problem and model its
knowledge base, it is fairly easy to apply related problem solving technique on this
problem without significant changes. However, one of the shortcomings of GTs is that
it is very difficult to reformat/ change if you need to update your system structure
(you might need to utilize a completely different GT with different problem solving
techniques for the new update).
Figure 2: Relations between Knowledge model components
Tasks- Goalsg
Problem Solving methodsg
GENERATEg
Task instancesg
INVOKEg
Inferencesg
REFER TOg
Ontologiesg
DESCRIBEg
Domain Knowledgeg

Another drawback for GTs is that they mainly focus on the knowledge part of the
system. More enhanced knowledge models have been designed to overcome this. We
will mainly focus on CommonKADS as one of these knowledge models.
b-CommonKADS
CommonKADS is a comprehensive methodology for KBS systems. It has been
gradually developed and has been validated by many companies and universities in
the context of the European ESPRIT IT Programme [2]. CommonKADS includes the
organizational system of the whole system. CommonKADS is also adaptive to
existence of intelligence within the system structure. The CommonKADS model is
shown in Figure 3.
Figure 3: The CommonKADS model
The concept partition resonates to the knowledge model described by GT. The context
represents the interface with the user. The artifact part shows the specifications and
requirements for the utilized platform, software modules, .. etc. Table 1 shows the
definition for each model. Relationship description between each model is shown in
Figure 4.
DefinitionModel
It describes and analyzes the main activities of an
enterprise.
Organization Model
It analyzes the organization's global subprocess scheme:
input, output, preconditions, performance criteria,
resources, and competencies.
Task Model
Agent characteristics description as task executors:
competencies, authorizations, and restrictions.
Agent Model
Conceptual description of agent transactions involved in
a task.
Communication Model
Description of knowledge types and structures used in a
task and the role of these knowledge components in the
task resolution, but implementation independent.
Knowledge Model
Starting with the previous models, this one describes the
technical specifications such as architecture,
implementation platform, software modules, etc., in
order to get the functionality specified in Knowledge and
Communication models.
Design Model

Table 1: Definitions of CommonKADS models
Figure 4: Relationship description between CommonKADS models
There are different knowledge categories within CommonKADS. These
categories are divided as:
 Task knowledge
o goal-oriented
o functional decomposition
o Controls inference knowledge
 Inference knowledge
o basic reasoning steps that can be made in the domain
knowledge and are applied by tasks
o Uses Domain Knowledge
 Domain knowledge
o relevant domain knowledge and information
o static
In order to build a structure for a specific system utilizing the CommonKADS model,
there are main steps that have to be performed. These steps can be summarized as:
1- Knowledge Elicitation
2- Identify domain knowledge
3- Identify task decomposition structure
4- Identify Inference knowledge model

Within the following section we will give a description of a traffic system, and how
we utilize CommonKADS for modeling.
Section 2: Traffic Systems
In most of the research related to modeling the knowledge base of a traffic system [3],
the system is identified as follows: a centralized traffic control center which
views/manages different feeds from cameras in the streets. There are also vehicle
detectors on the ground. According to the information captured from the cameras and
other devices, the centralized traffic control center takes decisions to control the
traffic. This system is shown graphically in figure 5.
Figure 5: Generic Traffic System
As mentioned in section 1(b), in order to build the traffic system knowledge model
according to CommonKADS, we need to follow these four steps: Knowledge
Elicitation, Identify domain knowledge, Identify task decomposition structure, and
Identify Inference knowledge model. Each step is described within the following sub-
sections.
Section 3: CommonKADS and Traffic Systems
Knowledge Elicitation
In [4], a re-usable elicitation method has been identified for traffic systems. In this
modeling context, an activity based on reusable experiences (i.e cases) was utilized.
In order to reach these cases, the following functions need to be applied on the
incoming data input:
1- Identify relevant descriptors of the incident case model.

2- Identify discriminant index to organize the case base.
3- Define a similarity metric for matching.
4- Register knowledge necessary to adapt solution part of the selected
case, in order to solve the current problem.
Elicitation sessions of the expertise are based on different methods: document
analysis, interviews, repertory grids, traffic manager's activities analysis and results of
the activity (problem reports). Each method presents an own goal and allows,
generally, to obtain a particular type of knowledge. Therefore, it is necessary to use
these methods concurrently, one cancelling out the drawbacks of others taken apart,
benefiting the qualities of each one.
From interviews with traffic management experts, the authors within the paper have
found that a large part of their knowledge is episodic. That is, the expert solves a new
problem by relating the current network situation to his previous experiences. These
experiences are sometimes specific incidents, with real dates and places, and
sometimes general classes of similar occasions.
Identify domain knowledge
In order to build a complete intelligent knowledge system, we have to identify the
main concepts/classes and functions to be done within a traffic system [5]. Figure 6
entails the domain knowledge concepts and its interaction with other models such as
the task and agent models.
Figure 6: Traffic system Domain layer and its interaction with knowledge
models.

On the other hand, Figure 7 details the main functionalities performed within a traffic
system (as described in [6]). We can see that at the sensors side, a video processing
kernel is included with the communication protocol with the camera devices. This
module is then followed by the methods done in order to achieve certain set of
statistics to be finally displayed to the user through the user interface as requested.
Figure 7: Traffic system Domain architecture
Identify task decomposition structure
There are three main composite tasks that we can identify to be happening in our
system: Diagnostic, prediction, and configuration tasks [3], all of which are re-
iterative and may occur in run-time. Each of these tasks is then decomposed into sub-
tasks until we reach the elementary non-composite tasks such as estimating the global
traffic demand (Figure 8).
Identify Inference knowledge model
Each of the before mentioned non-composite tasks (called an inference) has an input
and output role. They also utilize specific domain knowledge. The inference
knowledge model for the same example is shown in Figure 9.

Figure 8: Task decomposition structure
Although we have shown that CommonKADS has been used extensively in modeling
the knowledge base of traffic systems, some shortcomings have been discovered.
These shortcomings include:
* It focuses on human-computer not computer-computer interactions
* A restricted form of dynamic task assignment can be done
* Multi-partner transactions not dealt with naturally
Therefore, a new modification has been applied to CommonKADS to adapt to the
existence of new traffic systems following the multi-agent model, called Multi-agent
systems (MAS) CommonKADS.

Figure 9: Inference knowledge model
Section 4: MAS-CommonKADS
MAS-CommonKADS extends the knowledge engineering methodology
CommonKADS with techniques from object oriented and protocol engineering
methodologies. In order to adapt to the existence of multiple agents, a new model is
added to CommonKADS which is the "coordination model" [7, 8]. Figure 10 shows
the new model hierarchy.
Figure 10: MAS CommonKADS model hierarchy
In figure 11, we can see the detailed structure of the new added coordination model
and how it interacts with other models within the CommonKADS. We can see that the

coordination model relies on defining an organized communication window, where
each agent-agent communication happens through a certain session with a specific
protocol.
Figure 11: Coordination model relations
With the addition of the coordination model, there are certain knowledge categories
that need to be added to the standard CommonKADS knowledge types. Shown in
Figure 12 under the coordination knowledge hierarchy are the knowledge criteria
suggested by [3].
Research has been able to reach an adaptive knowledge model for traffic system
which realized the existence of multiple interacting agents as we have seen through
the report. However, current traffic systems has been effectively decentralized, where
crowd-sourced data made it possible for each user to change the organization structure
of the traffic system by representing a sensor model on their own [8]. As a result, an
addition is suggested within the following section.

Figure 12: Coordination knowledge
Section 5: Dynamic Virtual organization
A dynamic virtual organization (DVO) identifies the collaboration between partners
and the technological agility that are essential to the new trends of businesses [9], and
there exists research that worked extensively on building and automating the
knowledge model for the DVO creation [9]. In addition, one of the applications that
can extensively rely on DVOs is an intelligent traffic system [10]. Therefore, it is only
logical to incorporate the CommonKADS model for DVOs within the knowledge
model of an intelligent traffic system in order to take into consideration the variation
on the organization level.
Conclusions:
This report focuses on the incorporation of CommonKADS as a knowledge modeling
technique for traffic systems. We have been able to identify the main model and
analyze its shortcomings. Accordingly, we have identified related research that
worked on fixing similar issues and how it makes the suggested knowledge model
more adaptive to the current changes within traffic systems that is acting more and
more as an independent intelligent system. This report shows a great potential for
future work within the field of knowledge modeling traffic systems.

References:
1. Bylander, Tom, and B. Chandrasekaran. "Generic tasks for knowledge-based
reasoning: the “right” level of abstraction for knowledge
acquisition."International Journal of Man-Machine Studies 26.2 (1987): 231-243.
2. Wielinga, Bob, et al. "The CommonKADS framework for knowledge
modelling."Knowledge Acquisition for Knowledge Based Systems Workshop.
1992.
3. Molina, Martin, J. O. S. E. F. A. HERN Á, and JOS É. CUENA. "A structure of
problem-solving methods for real-time decision support in traffic
control."International Journal of Human-Computer Studies 49.4 (1998): 577-600.
4. Caulier, Patrice, and Bernard Houriez. "A case-based reasoning approach in
network traffic control." Systems, Man and Cybernetics, 1995. Intelligent Systems
for the 21st Century., IEEE International Conference on. Vol. 2. IEEE, 1995.
5. Dieng, Rose, et al. "Building of a corporate memory for traffic-accident
analysis." AI magazine 19.4 (1998): 81.
6. Abreu, Bruno, et al. "Video-based multi-agent traffic surveillance
system."Intelligent Vehicles Symposium, 2000. IV 2000. Proceedings of the IEEE.
IEEE, 2000.
7. Iglesias, Carlos A., et al. "A methodological proposal for multiagent systems
development extending CommonKADS." Proceedings of the 10th Banff
knowledge acquisition for knowledge-based systems workshop. Vol. 1. 1996.
8. Gascuena, Jose M., and Antonio Fernández-Caballero. "On the use of agent
technology in intelligent, multisensory and distributed surveillance." Knowledge
Engineering Review 26.2 (2011): 191-208.
9. Yassa, Morcous. "Utilizing CommonKADS as Problem-Solving and Decision-
Making for Supporting Dynamic Virtual Organization Creation." IAES
International Journal of Artificial Intelligence (IJ-AI) 3.1 (2014).
10. H. Afsarmanesh, “Virtual Organizations in a Dynamic Context”, Lecture notes,
Federated collaborative network group, University of Amsterdam, 2012

Report on Knowledge Modeling in Various applications in Traffic Systems

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Report on Knowledge Modeling in Various applications in Traffic Systems (20)

More from Yomna Mahmoud Ibrahim Hassan (15)

Recently uploaded (20)

Report on Knowledge Modeling in Various applications in Traffic Systems