Anticipatory scheduling of synchromodal transport using approximate dynamic programming

Abstract We study the problem of scheduling container transport in synchromodal networks considering stochastic demand. In networks, transportation modes can be selected dynamically given actual circumstances and performance is measured over entire network time. model this as a Markov Decision Process propose heuristic solution based on Approximate Dynamic Programming (ADP). Due to multi-period nature problem, one-step look-ahead perspective traditional approximate value-iteration approach make flounder end local-optimum. To tackle this, we inclusion Bayesian exploration using Value Perfect Information (VPI). series numerical experiments, show how VPI significantly improves ADP algorithm. Furthermore, our proposed ADP–VPI combination achieves significant gains common practice heuristics.