Computation of an Optimal Control for a Signalized Traffic Intersection

A model of a signalized intersection is described in which the development of the queues in time is described by a controlled Markov chain with the state of the light as the control variable. By the method of dynamic programming it is possible to compute a control minimizing the expected value of a suitable cost function. This control becomes a function of the queue lengths. The result of the computation for a model example is presented.