Evolutionary Algorithms for Optimizing Traffic Signal Operation

Maintaining efficient transportation systems has become a key goal for government-based transportation agencies in this age of serious energy shortages, rising costs and competing recipients of public funds. The emphasis has shifted to optimizing traffic movement on the existing surface street network rather than to provide more freeways. As described in Schnabel and Lohse (1997), the network capacity is limited by intersections. In this paper, we present an approach to automated optimization of traffic signal operation. We present a problem decomposition and encoding for the use of different traffic models as well as a variety of optimization tools. First experimental results show that an automated design and optimization of intersection signal operations is not only possible but yields good results. This paves the way for on-line self-adapting signal operation: combined with a traffic observation system it is not only possible to react to short-term variations in traffic flow but also to adapt to long-term changes of the situation.