Congestion due to traffic accidents and incidents can be reduced through effective freeway incident management. However, this is plagued by a number of problems and requires a high level of expertise and coordination among the involved personnel. The ill-structured nature of the problem, constantly changing conditions, the number of agencies involved, and the lack of current information often cause errors in decision and response. Under these conditions, there is need for computer based support tools to provide the required decision and information support and aid the entire process by improving coordination and communication.

This study focuses on addressing this issue through the development of an Expert-GIS system which integrates the powerful spatial data handling capabilities of a Geographic Information System with the rule based reasoning logic of an Expert System. The system is designed as a Group Decision Support System that provides the required support for both the substance of the problem (decisions) and the agency level interactions that take place. The ability to support the process of response is modeled using a blackboard architecture for the system.

The prototype developed fully integrates the software environments of Arc/Info and Nexpert-Object and presents a unified interface, from where different incident management functions can be accessed. A complete spatial database was designed for the Fairfax County in Northern Virginia as a part of this development effort. Decision support is provided through a set of six integrated modules - incident detection and verification, preliminary response, duration estimation, delay calculation, final response plan and diversion planning, and recovery. Coordination and communication were enhanced by ensuring the uniformity of information at different locations using the system, and through a messaging mechanism that informed users about the current status of incident.

The prototype system was developed for two hypothetical agencIes called the Traffic Management Center and The Police Control Center. Historical incident cases were use to test these systems and check the accuracy of the database and the rule base. Both the tests and the development effort showed a strong need for established sources of network information, that could be readily incorporated into the database.

Given the fact that the system works with real network data, the next phase of research in should focus on the deployment of the system at test sites. User feedback obtained from these tests would then serve as a basis for future enhancements.