ABSTRACT
In this paper we present a multiagent environment suitable to be applied on a system for controlling urban-traffic. This work is an evolution of the expert system prototype developed inside the ESPRIT II EQUATOR project[EQ94], which incorporates reasoning frameworks of the application domain theories by means of a qualitative/temporal model of the system behaviour. The expert system is composed by six functional agents which respectively perform pre-processing of data, detection of current problems, diagnosis of problems, short term prediction, generation of solutions and refinement of solutions. For each one of the system objects (streets, intersections, etc.), the temporal knowledge base (TKB) contains the past, present and future events that are important in their own evolution. This paper describes the structure of this TKB and proposes a new and efficient, but application oriented, heuristic algorithm for doing the reasoning maintenance process.
