A stochastic programming approach for robust vehicle scheduling in public bus transport

We present a new stochastic programming approach for robust vehicle scheduling in public bus transport. Our approach uses typical disruption scenarios during the optimization to minimize the expected sum of planned costs and costs caused by disruptions. The schedule is represented as a time-space network with all connecting arcs to enable independent penalization of every connection between two consecutive service trips. Our method significantly decreases total expected costs compared to just minimizing planned costs and outperforms a simple approach of adding fixed buffer times between service trips. Despite the increased computational complexity, small and medium-sized real-world instances can be solved.