Optimal Planning on a Single-Route Transit System with Modular Buses

This paper considers optimal planning of modular buses on a single route to minimize the average passenger travel time. Modular vehicles consist of individual units that can attach and detach from other modules at any stop, allowing for increased flexibility. As the passenger demand and the travel time between stops are stochastic variables, we use a Markov decision process to model the system. The problem of minimizing passenger travel time is converted to optimally planning the actions of all the vehicles, which include stopping at a stop, skipping the stop, and attaching or detaching from other modules. We develop a cost formulation to capture the impact of these actions on passenger travel time, and propose a control policy that selects an action with the minimum cost. To evaluate our proposed policy, we compare it in simulation with two other policies: when vehicles serve every stop without attaching or detaching, and a previously proposed rule-based attaching/detaching policy. Our proposed policy significantly outperforms both of these strategies.