Neural network models for traffic control and congestion prediction

Advance Traffic Management Systems (ATMS) must be able to respond to existing and predicted traffic conditions if they are to address the demands of the 1990s. Artificial intelligence and neural network are promising technologies that provide intelligent, adaptive performance in a variety of application domains. This paper describes two separate neural network systems that have been developed for integration into an ATMS blackboard architecture. The first system is an adaptive traffic signal light controller based upon the Hopfield neural network model, while the second system is a backpropagation model trained to predict urban traffic congestion. Each of these models is presented in detail with results attained utilizing a discrete traffic simulation shown to illustrate their performance.