We present the optimization component of a decision support system developed for a sedan service provider. The system assists supervisors and dispatchers in scheduling driver shifts and routing the fleet throughout the day to satisfy customer demands within tight time windows. We periodically take a snapshot of the dynamic data and formulate an integer program, which we solve to near-optimality using column generation. Although the data snapshot is stale by the time a solution is computed, we are able to solve the integer program quickly enough that the solution can be adopted after minor modifications are made by a fast local-search heuristic. The system described in this paper is currently in use and has improved the provider's productivity significantly.
IBM Research Report RC23208