Service and transfer selection for freights in a synchromodal network

We study the problem of selecting services and transfers in a synchromodal network to transport freights with different characteristics, over a multi-period horizon. The evolution of the network over time is determined by the decisions made, the schedule of the services, and the new freights that arrive each period. Although freights become known gradually over time, the planner has probabilistic knowledge about their arrival. Using this knowledge, the planner balances current and future costs at each period, with the objective of minimizing the expected costs over the entire horizon. To model this stochastic finite horizon optimization problem, we propose a Markov Decision Process (MDP) model. To overcome the computational complexity of solving the MDP, we propose a heuristic approach based on approximate dynamic programming. Using different problem settings, we show that our look-ahead approach has significant benefits compared to a benchmark heuristic.