SYSTEM ARCHITECTURE OF THE DECISION SUPPORT

SYSTEM EMPLOYING MICROSCOPIC SIMULATION AND

EXPERT SYSTEM IN PARALLEL FOR THE POST INCIDENT

TRAFFIC MANAGEMENT

S. Akhtar Ali Shah

and Hojung Kim

Institute of Geography, Urban and Regional Planning, University of Peshawar, Peshawar, Pakistan

L.G. Electronics, Seoul, Republic of Korea

Keywords: Traffic management center, Micro simulation, Expert system.

Abstract: This paper presents system architecture of the post-incident decision support system (PIDSS), which

incorporates the predicted incident impacts from an offline microscopic simulation platform into an expert

system. The system yields an immediate operational strategy for the freeway managers that can further be

fine-tuned with the online simulation results. The novel idea presented in this paper is the replacement of

the domain expert and knowledge engineer with the output of the microscopic simulation that would make

post incident congestion mitigation on the road network more efficient and cost effective.

1 INTRODUCTION

Non-recurring congestion is the result of traffic

accidents, bad weather, road works or unplanned

special event that disrupts traffic flows and causes

unexpected delays. It abruptly reduces the available

capacity and reliability of the entire transportation

system and thus needs to be intelligently tackled.

.The orthodox approach of tackling the non-

recurring congestion relies heavily on the individual

expertise and experience of the traffic managers who

tend to nominate certain heuristics (or guesswork)

about the impacts of an incident. This approach may

cause time loss and induces inconsistency in the

entire incident management operation resulting in an

inefficient use of resources and uncoordinated

mitigation strategy.

In a post-incident scenario, the problems faced

by the traffic managers at the Traffic Management

Center (TMC) have also been acknowledged by

Nagel and Schreckenberg (1992) and there have

been a significant number of other attempts in this

area by researchers as well as practitioners (Hayashi,

Morisugi, 2000; Yoon, et al., 2008). However initial

algorithms for the incident related congestion

mitigation were mainly based on the analytical

models, knowledge-based expert system and

geographic information system (GIS). (Xaichen and

Daniel, 1995) used cellular automata (CA) model for

the prediction of flows using real-time inductance

loop data for freeway traffic that demonstrates the

viability to integrate inductance loop data, cellular

automata and car-following models to simulate the

traffic dynamics for the prediction of the post

incident traffic flows.

This article imparts the system architecture of a

hybrid solution that is named as post-incident

decision support system (PIDSS) that employs

micro-simulation in conjunction with intelligent

system for the analysis based traffic management.

The cornerstone of the approach is an appreciation

of the real incident scenario that demands an

expeditious decision and the participation of the

local network managers for the optimal effectiveness

and coherence through automation of the whole post

incident decision-making process. The feasibility

aspects and specification of requirements of PIDSS

are discussed along with its ability to work in post-

incident scenario for the efficient functioning of the

freeways as well as the urban road networks.

113

Akhtar Ali Shah S. and Kim H..

SYSTEM ARCHITECTURE OF THE DECISION SUPPORT SYSTEM EMPLOYING MICROSCOPIC SIMULATION AND EXPERT SYSTEM IN

PARALLEL FOR THE POST INCIDENT TRAFFIC MANAGEMENT.

DOI: 10.5220/0003385301130117

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 113-117

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2 THE CONCEPTUAL

FRAMEWORK OF PIDSS

The conceptual framework of PIDSS is based on

associating the output of the freeway incident

analysis system (FIAS), in the whole process of

traffic management and modifying the mitigatory

measures as per online simulated results (Figure1).

The FIAS, which was developed earlier by the

authors and has been, discussed elsewhere in detail

(Kim et al., 2004 and Shah et al. 2008) employs

historical data supplemented with the real-time data

from the toll collection system (TCS) and the

vehicle detection system (VDS) as well as the spatial

data on micro-simulation platform.

The perception of an incident and its impact

forecasting is the kernel in the whole process of non-

recurrent congestion management and needs to be

consistently measured with a significant degree of

reliability. In this context, findings of the micro

simulator (FIAS) are found useful and can be

injected as a replacement to the traditional heuristics

that are neither consistent nor tangible. As online

simulation needs certain time before displaying the

impacts of an incident and the scenario requires

immediate mitigatory measure, therefore, offline

simulation results are used to devise an immediate

mitigatory plan that shall be refined and updated

once real-time incident impact data is available

(Figure 1).

3 KNOWLEDGE-BASED EXPERT

SYSTEM

In a post-incident scenario, a knowledge-based

expert system (KBES) has been conventionally

Figure 1: A conceptual framework of PIDSS.

recommended because of its potentials to streamline

and automat certain low level procedural tasks and

emulate the experienced traffic manager (Zhang and

Ritchie 1994; Ritchie, 1990; Flippo and Ritchie,

2002). The essence of these systems is the

application of expert’s knowledge in a narrow and

well-defined problem arena. A KBES can employ

symbolic reasoning and heuristics in problem

solving that enhances its suitability for the complex

scenario analysis with deficient algorithmic solution

(Mitrovich, et al., 2006).

Santa Monica Freeway Smart Corridor (Flippo

and Ritchie, 2002) project was one of the very first

attempt in which a full-fledged real-time knowledge-

based expert system was employed for the traffic

surveillance and control purposes. In this work, a

conceptual framework of a multiple real-time

knowledge-based expert system was suggested.

Zhang and Ritchie (2004) propose a knowledge-

based expert system and name it as freeway real-

time expert system demonstration (FRED). Their

proposed methodology is to follow experienced

TMC operators’ and traffic engineers’ approach

using an expert system. It incorporates symbolic

reasoning and heuristics for solving the problem.

This system is capable of network level operations,

multiple incident handling, better user interface and

incident recovery monitoring. Nonetheless, FRED

does not predict the post-incident traffic delays and

the significance of the historic data are also not

realized in the analysis.

In another attempt (op cit.) a knowledge-based

system is employed focusing on the cooperative

inter-jurisdictional traffic management. The system

adopts a multi-decision maker approach that reflects

the spatial and administrative organization of traffic

management agencies in US cities. It provides a

cooperative solution that exploits the willingness of

agencies to cooperate and unify their problem

solving capabilities without compromising their

individual authority and the inherent distribution of

data and expertise.

4 SYSTEM ARCHITECTURE

The indispensable part of modelling is to establish a

logical information exchange between the user and

the system besides in vitro processing and

simulation methods that takes place within the

system. System architecture constitutes interactions

amongst various components of the system for

achieving predefined user objective. The two

elements of system architecture of PIDSS are

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

114

described in greater depth below.

4.1 Logical Architecture

The logical architecture of PIDSS (Figure 2)

primarily considered as a ‘specification of

requirement’ upon which the working model of

deployment guide exists. The main activity flow

diagram is flanked with two fundamental units of the

architecture: FIAS and the Simulation Based Expert

System (SBES). Incident detection (using prevailing

algorithms like speed map) and verification is

followed with the insertion of incident parameters

like type, severity, cross sectional location, time,

number of vehicle involved in the expert system.

The domain of the expert system extends from

incident categorization based on input parameters;

through impact assessment for encapsulating each

individual category of incident; to the application of

different mitigatory measures. These tools are

mostly ITS applications that are primarily operation

and information oriented at local level but may

cause significant travel and traffic impacts at a

regional levels.

Autonomously, managers also instigate FIAS for

the cross reference and any subsequent tuning in the

impact assessment is based on online simulation in

the later stages of operation. SBES also contains a

significantly large data bank of offline coordinated

relative weight of individual measures and its linear

combination for the optimal mitigatory impacts. The

relative weights are derived from the significance of

each measure and its overall performance and

impacts on the operational functioning of the whole

network.

The coordinated initial mitigatory plan is thus

injected into the system that has been locally

optimized and is reviewed for any changes in

consultation with the lower tier of the system

network at local level to avoid any friction from

other stakeholders in the system before effectuation.

This pluralistic approach is essential and highly

recommended to use the full potential of the

physical infrastructure. The system also provides a

real-time monitoring mechanism that suggests

improvement in a running mitigatory plan or any of

its components. This is supplemented with a post-

incident evaluation that helps in the expert system

database evolution.

Incident detection

Verification

Incident parameters

Incident categorization

Impacts assessment

Mitigatory plan

formulation

Plan review and

revision, if needed

Execution of the final

mitigatory plan

Performance

Monitoring

Post incident

Evaluation

Offline

simulation

FIAS

Online

simulation

FIAS predictions of

the impact

Revision

if needed

Local traffic

management center

ES upgrading

SBES

- Categorization

- Impact assessment

- Control measure assessment

- Coordination module

Incident detection

Verification

Incident parameters

Incident categorization

Impacts assessment

Mitigatory plan

formulation

Plan review and

revision, if needed

Execution of the final

mitigatory plan

Performance

Monitoring

Post incident

Evaluation

Offline

simulation

FIAS

Online

simulation

FIAS predictions of

the impact

Revision

if needed

Local traffic

management center

ES upgrading

SBES

- Categorization

- Impact assessment

- Control measure assessment

- Coordination module

Figure 2: The logical framework of PIDSS.

4.2 Physical Architecture

The physical architecture of PIDSS (Figure 3)

addresses organization of the system on the

functional lines which is based on information flow

mapping and logical modeling. It supports a range of

evaluation conditions and effectuation strategies in

terms of standardized framework of the logical

architecture. The architecture generates information

in a format that is more tangible and can invoke a

locally optimized and globally integrated mitigatory

plan at its terminal point.

Traffic managers are placed at the crux level and

have interfacing with every internal subsystem and

execute a control over all data flow. The essential

architectural flows in the key logical units like

incident categorization; impact assessment or

strategy coordination with the local traffic mangers

encapsulates huge data sharing and information

exchange between the terminal point and the

subsystem. The manger, who has access to the

essential ITS infrastructure like the surveillance

system; verification infrastructure; FTMS and

highway geographic information system (HGIS)

server, injects incident parameters into the Expert

System

4.3 Simulation based Expert System

Collecting knowledge needed to solve problems and

build the knowledge base continues to be the biggest

bottleneck in building expert systems. This

SYSTEM ARCHITECTURE OF THE DECISION SUPPORT SYSTEM EMPLOYING MICROSCOPIC SIMULATION

AND EXPERT SYSTEM IN PARALLEL FOR THE POST INCIDENT TRAFFIC MANAGEMENT

115

impedance is resolved in this hybrid system by

replacing an expert domain of a conventional expert

system with the microscopic simulation platform

trained in knowledge acquisition and representation

into a dependency network. The high level structure

around the fundamental activities of the system

defined in the logical architecture is realized in an

orthodox transport planning paradigm with two

typical modules: analysis and intervention. The

analysis stage is further sub classified into main

categories incident categorization and impact

assessment module. The intervention stage

encompasses coordination (with the local agency)

module and mitigatory measure module. The key

modules of the stages are discussed below.

4.3.1 Incident Categorization Module

This is an ES tool that provides a representation

scheme for incident categorization expressing

knowledge about parameters and categories.

Parameters include location, type and severity,

number of vehicles involved, critical section and

capacity reduction, estimated duration, and location

of the nearest rescue infrastructure. These

parameters classify the incident into previously

defined categories as per type and combination of

parameters.

4.3.2 Impact Assessment Module

This module is one of the indispensable components

of the proposed expert system. It replaces the

conventional knowledge base with both off and on-

line versions of FIAS impacts of all predefined

categories, which are assessed and tabulated as a

knowledge base. The FIAS off line simulation

module is a major component in the knowledge base

that encapsulates all categories of an incident for

impacts prediction and subsequent indexing. On-

line FIAS predictions bring dynamism in the system

with the supplementation of real-time scenario.

4.3.3 Coordination Module

Several research development and deployment

projects have acknowledged the importance of

multi-agent, coordinated and inter-jurisdictional

approach that was reported by Flippo and Ritchie,

(2002). The multi-decision making in PIDSS

simplifies the scenario by executing a knowledge

base for the priorities of the local authority of the

abutting road network regarding different mitigatory

measures and its diversity with the many other

parameters like category, type and location of the

incident and its impacts on the general flows in the

area. These knowledge bases shall be reviewed

periodically and updated by assigning the revised

priority through relative weighting system approach.

Besides the real-time participation of the local

agencies is also insured using a fast TCP/IP based

communication protocol.

4.3.4 Mitigatory Plan Formulation Module

Mitigatory plan can be classified into two groups:

Preliminary and revised. The preliminary plan

originates from the knowledge-based relative

weighting system of individual mitigatory operation

selected by the respective stakeholders from the

local areas as well as the freeway agencies. However

the dynamism of the system allows for any real-time

incorporation of the changes in the values of the

relative weighting depending upon the revised

priorities for any local reason like time of the day,

conditions of the network and so on. These revisions

are amalgamated into the plan, which will be

regarded as the revised plan and executed. The real-

time coordination with the local agents allows for

the optimization and globalization of a solution

using the full potentials of the network as a unit.

Local area

traffic manager

Rescue

infrastructure

Traffic manager

Surveillance

system

Verification

system

Incident

Parameter

FIAS

FTMS server

HGIS

SBES

Incident categorization

module

Impact assessment

module

Coordination

module

Mitigatory

module

Incident parameters

Categorization

knowledge bank

Impacts on local traffic

Capacity of the local

network

Off-line FIAS operation

and building of

knowledgebase for

impacts

Impacts of individual

measures

Combined impact

Selection of the

optimum combination

FIAS online

simulation and

prediction of impacts

Comparison with the

database impact and

fine turning if needed

Execution of the

selected optimum

mitagory plan

Monitoring and fine

turning if needed

Local area

traffic manager

Rescue

infrastructure

Traffic manager

Surveillance

system

Verification

system

Incident

Parameter

FIAS

FTMS server

HGIS

Traffic manager

Surveillance

system

Verification

system

Incident

Parameter

FIAS

FTMS server

HGIS

SBES

Incident categorization

module

Impact assessment

module

Coordination

module

Mitigatory

module

Incident parameters

Categorization

knowledge bank

Impacts on local traffic

Capacity of the local

network

Off-line FIAS operation

and building of

knowledgebase for

impacts

Impacts of individual

measures

Combined impact

Selection of the

optimum combination

FIAS online

simulation and

prediction of impacts

Comparison with the

database impact and

fine turning if needed

FIAS online

simulation and

prediction of impacts

Comparison with the

database impact and

fine turning if needed

Execution of the

selected optimum

mitagory plan

Monitoring and fine

turning if needed

Execution of the

selected optimum

mitagory plan

Monitoring and fine

turning if needed

Figure 3: Physical architecture of PIDSS.

5 DISCUSSIONS AND

CONCLUSIONS

This paper discusses an application of transport

telematics to automate decision support for the TMC

personnel in the post-incident scenario using a

hybrid system of microscopic simulation and

artificial intelligence. The unique feature of PIDSS

is the immediate decision making in the crises state

incorporating offline simulation results used to

instigate a preliminary mitigatory plan. The tool

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

116

facilitates local integration for a regional solution

incorporating the predefined interest of all key

stakeholders employing relative weighting

algorithm, which is rapidly coordinated and updated

using IT infrastructure. This initial plan is revised

with the FIAS online simulation for incidents

impacts prediction that incorporates real data with

historical and spatial data in the incident analysis

stage. The flexibility of the tools allows for the

incorporation of the priority revision of all

stakeholders with mutual understanding and

collaboration.

In the development of PIDSS, unlike some of the

known algorithms (Zhang and Ritchie 1994; Ritchie,

1990; Flippo and Ritchie, 2002), a different

approach of the knowledge acquisition was opted.

The domain expert was replaced with the

microscopic simulation platform, trained in

knowledge acquisition and representation. Using a

data manipulation algorithm the outputs of

simulation are transformed into dependency

networks (an outline of the rules), which is

subsequently coded and programmed into the

system. Thus the simulation replaces both the expert

domain (the source of knowledge) and the designers

of the expert system (knowledge engineer). The

most obvious advantage of this development method

is its cost effectiveness to build expert systems to

eliminate the need for an expert domain and the

knowledge engineer for the extraction and

representation of knowledge. Nonetheless the crucial

advantage is the speed, coordination and time saving

in a crisis scenario.

PIDSS knowledge base is rooted into a

microscopic simulation based model that predicts

the post-incident traffic impacts, which is imperative

for the real-time incident analysis and improves the

functioning of TMC. It is anticipated that the

incident analysis result in this format will help the

traffic managers to take significantly consistent steps

based on tangible information and not the

speculative approach.

ACKNOWLEDGEMENTS

Part of this paper was developed in collaboration

with the Korea Highway Corporation (KHC).

However its content reflects views of the authors; it

neither constitutes a standard, specification or

regulation nor official views or policy of the KHC.

REFERENCES

Flippo Loggi, Ritchie S. G., 2002 Multi-agent architecture

for cooperative inter-jurisdictional traffic congestion

management Transportation Research – Part C: -

Emerging Technologies Volume 10. Issue 5–6, pp.

507-527, Elsevier Science Ltd, Oxford, England

Hayashi, H. and Morisugi, H., 2000, International

comparison of background concept and methodology

of transportation project appraisal, Transport Policy 7,

No. 1, pp. 73-88,

J. K. Kim, R. Sharman, H. R. Rao and S. Upadhyaya,

2007, Efficiency of critical incident management

systems: instrument development and validation,

Decision Support Systems Volume 44, Issue 1 Elsevier

Science Bv Amsterdam, Netherlands

Kim Ho-Jung, S. Akhtar Ali Shah, Seungkirl Baek, Ahn,

Byung Ha, 2004, Architecture of decision support

system for freeway incident analysis using

microscopic simulation –A case study of Republic of

Korea. Proc. of 11

World Congress on ITS Nagoya,

Aichi Japan (CD ROM).

Nagel K. and Schreckenberg M. 1992. A Cellular

Automaton Model for Freeway Traffic J. Physique I,

vol. 2, Issue 12 France pp.2221-2229

Ritchie S. G. 1990 A knowledge-based decision support

architecture for advanced traffic management

Transportation Research Part A: Policy and Practice,

vol. 24, no. 1, pp. 27-37, Pergamon-Elsevier Science

Ltd, Oxford, England

S. Akhtar Ali Shah; Hojung Kim; Seungkirl Baek;

Hyunho Chang; Byung Ha Ahn 2008, System

Architecture of A Decision Support System for

Freeway Incident Management in Republic of Korea,

Transportation Research Part A: Policy and Practice,

Volume 42A, Pergamon-Elsevier Science Ltd Oxford,

England

S. W. Yoon, J. D. Velasquez, B. K. Partridge and S. Y.

Nof, 2008 Transportation security decision support

system for emergency response: A training prototype

Decision Support Systems, Volume 46, Issue 1,

Elsevier Science Bv Amsterdam, Netherlands

Sergio Mitrovich, Gaetano Valenti, Massimo Mancini,

2006, A Decision Support System For Traffic Incident

Management, Roadway Tunnel Infrastructure

Association for European Transport and contributors

<website> www.etcproceedings.org/paper/download/

1414 (accessed Nov 2009)

Xaichen Xu, Daniel J. Dailey, 1995, Real-time Traffic

Simulation and Prediction Using Inductive Loop Data

Proc. of the Vehicular Navigation and Information

Systems, Seattle, Washington, pp.194-199,

Zhang H. and Ritchie S. G. 1994, Real-time decision-

support system for freeway management and control

Journal of Computing in Civil Engineering, vol. 8, no.

1,pp. 35-51, American Society of Civil Engineers

Reston, Virginia, USA

SYSTEM ARCHITECTURE OF THE DECISION SUPPORT SYSTEM EMPLOYING MICROSCOPIC SIMULATION

AND EXPERT SYSTEM IN PARALLEL FOR THE POST INCIDENT TRAFFIC MANAGEMENT

117