Knowledge Based Approach To Urban TrafficControl

We present the general design of a system to control urban street traffic signals. It is based on cooperating, learning, realtime, distributed expert systems. We also describe the operation of a running prototype program which, while using several simplifying assumptions, has proven the technical feasibility of the approach. It has also attained a 36% improvement in the traffic flow under non-saturated conditions. Current developments include the design of a general-purpose system that can be customized to most street configurations. Finally, we draw conclusions concerning the distributed planning and problem solving methodology.