A neural network application in signal timing control

Generally the most cost-effective means of achieving optimized vehicle flow through a given road network is by improving the timing of traffic signals at network intersections. This paper uses neural networks (NN) as the bases for the control law. The NN weight estimation occurs real time in closed-loop mode via the simultaneous perturbation stochastic approximation algorithm. The approach results in a net 10-percent reduction in vehicle wait time over the performance of the existing, in-place strategy.