TRIAL-AND-ERROR IMPLEMENTATION OF THE CONGESTION PRICING ON NETWORKS WITH UNKNOWN DEMAND FUNCTIONS

Recent advances in congestion pricing studies reveal that the marginal-cost pricing tolls can be identified from traffic counts in an iterative trial-and-error procedure, without resorting to the predetermined analytical demand functions, which are generally difficult to estimate in practice. The theoretical foundation of such procedure has been established rigorously for a general network in the first-best pricing environment under which each link of the network is tolled. The perfect first-best pricing policy is, however, a rather theoretical, and the second-best issues instead have received ample attention in the recent literature. This paper makes a significant step towards efficient implementation of the trial-and-error procedure for the practical second-best pricing problem where only a subset of links is tolled and only a subset of link traffic counts is available in each trial. The objective is achieved by virtue of advanced bi-level programming techniques