Induction of fuzzy rules for a distributed traffic signal timing system

A distributed, fuzzy traffic signal timing system is described that utilizes independently controlled traffic signals where each signal employs a fuzzy rule tree to adapt its timing. Fuzzy rules capture the human reasoning ability that is used by a police officer directing traffic. In addition, a system of independent controllers may be capable of providing the benefits of an area-wide traffic control system to communities which can not afford the cost of the centralized control system, commonly in use today. A Simulation tool kit that has been combined with a fuzzy/inductive/adaptive expert system controller is described that can be used to model alternative traffic signal timing system concepts. System evaluation determines the required knowledge sharing among the independent controllers needed to achieve optimal area-wide system performance. Induction is used to generate a fuzzy rule tree for each alternative definition of knowledge sharing evaluated. Given each fuzzy rule tree, a feedback scheme is employed that measures the effectiveness of a decision with respect to the system goals and rewards or punishes the supporting rules, optimizing area-wide traffic control performance.<>