Operationalizing modular autonomous customised buses based on different demand prediction scenarios

This paper presents a novel framework for customised modularbus systems that leverages travel demand prediction and modu-lar autonomous vehicles to optimise services proactively. The pro-posed framework addresses two prediction scenarios with differ-ent forward-looking operations: optimistic operation and pessimisticoperation. A mixed integer programming model in a space-time-state network is developed for the optimistic operation to determinemodule routes, schedules, formations and passenger-to-moduleassignments. For the pessimistic case, a two-stage optimisation pro-cedure is introduced. The first stage involves two formulations (i.e.,deterministic and robust) to generate cost-saving plans, and thesecond stage adapts plans with control strategies periodically. ALagrangian heuristic approach is proposed to solve formulations effi-ciently. The performance of the proposed framework is evaluatedusing smartcard data from Beijing and two state-of-the-art machinelearning algorithms. Results indicate that the proposed frameworkoutperforms the real-time approach in operating costs and high-lights the role of module capacity and time dependency.