Joint optimization of modular autonomous vehicles scheduling and seat allocation for reservation-based travel

This study investigates the joint scheduling and seat allocation problem for modular autonomous vehicles (MAVs) in a reservation-based transit system. Unlike traditional buses, MAVs can dynamically couple or decouple at stops, enabling flexible vehicle formations to match spatiotemporal demand. This study develop a mixed-integer linear programming (MILP) model to simultaneously determine departure schedules, stop-wise MAV configurations, and reserved seat allocations. To handle the computational complexity, a customized decomposition framework is proposed, which separates scheduling decisions in the master problem from seat allocation in the subproblem. Based on this framework, a hybrid decomposition variable neighborhood search (HDVNS) algorithm with a simulated-annealing-based acceptance rule is designed to efficiently solve large-scale instances. Numerical experiments consistently indicate that integrating reservation-based seat allocation with flexible MAV operations yields substantial improvements in system performance. In a small-scale example, incorporating seat-level assignment improves capacity utilization to 97.43% and reduces total system cost by 16% relative to a baseline without seat allocation. Large-scale case study further show that fully flexible MAV operations outperform fixed-capacity and partially flexible strategies, achieving up to a 41.98% reduction in overall system cost. The results highlight the potential of integrating MAV technology with reservation-based planning to enhance the adaptability and efficiency of urban transit systems.