Heuristic Decision Policies for the Control of Reversible Traffic Links

Mathematical models of the delays incurred in a traffic link with reversible lanes, as a function of demands for service and allocation decisions, are constructed assuming that the state of the operation is reviewed periodically. There is a penalty associated with each allocation change that is manifested as a loss of capacity. Heuristic algorithms are derived that enable the determination of decision policies that minimize the delay per vehicle during some horizon. The first phase of the determination is performed upon a basic model, in which the reversal penalties are ignored, and results in a sequence of optimal periodwise decisions. This policy is then subjected to a second phase, performed upon the lost time model, in which reversal penalties are included. Progressively better policies are generated in this phase according to the heuristic notion that an optimal policy does not differ much from the policy obtained in the first phase. The solution of the models is illustrated in a case study using actual data for the Chesapeake Bridge.