Traffic responsive signal timing plan generation based on neural network

This paper proposes a neural network based traffic signal controller, which eliminates most of the problems associated with TRPS mode of the closed loop system. Instead of storing timing plans for different traffic scenarios, which requires clustering and threshold calculations, the proposed approach uses an ANN model that produces optimal plans based on optimized weights obtained through its learning phase. Clustering in closed loop system is root of the problems and has been eliminated in the proposed approach. The particle swarm optimization technique has been used both in the learning rule of ANN as well as generating training cases for ANN in terms of optimized timing plans based on highway capacity manual delay for all traffic demands found in historical data. The ANN generates optimal plans online for the real time traffic demands and is more responsive to varying traffic conditions.