Multiobjective Model Predictive Control Applied to Dial-a-Ride System

A multiobjective model-based predictive control approach is presented for solving a dial-a-ride problem. The dynamic objective function considers two dimensions: user and operator costs. As those two components are usually aimed at opposite goals, the problem is formulated and solved through multiobjective model predictive control. When a new call asking for service is received, the method first solves a multiobjective optimization problem, providing the Pareto optimal set. Note that from this set just one solution has to be applied to the system. Then, the dispatcher participates in the dynamic routing decisions by expressing his/her preferences in a progressively interactive way, seeking the best trade-off solution at each instant among the Pareto optimal set. The idea is to provide to the dispatcher a more transparent tool for the decisions. Several criteria, emulating different dispatchers, are proposed in order to systematize different ways to use the information provided by the dynamic optimal Pareto front. An illustrative experiment of the new approach through simulation of the process is presented to show the potential benefits in the operator cost and in the quality of service perceived by the users.