On-Line Distributed Model Predictive Scheduling for Multi-Vehicle Routing Problems With Lane-Change and Platoon Maneuvers

Autonomous vehicles (AVs) have garnered significant attention in the field of smart logistics due to their potential to greatly enhance transportation efficiency. While significant focus has been placed on the capabilities of individual AVs, less attention has been given to the complexities of coordinating multiple vehicles within dynamic traffic environments. Key factors such as combined traffic flow, varying traffic signals, and complex cooperative maneuvers play a major role in effective multi-vehicle scheduling. This paper aims to address these overlooked challenges by exploring dynamic traffic conditions and micro-level cooperation within multi-vehicle routing systems. The large-scale multi-vehicle routing problem (MVRP) is reformulated into a series of smaller-scale cooperative vehicle routing problems (VRPs). A distributed model predictive control-based (DMPC-based) planner is constructed and applied in parallel across connected autonomous vehicles (CAVs). Each instance of distributed model predictive scheduling incorporates a closed-loop vehicle dynamics to predict lane-changing and platooning maneuvers, where feedback controllers are introduced. To optimize routes toward destination, a destination-oriented search space is rebuilt with the filtered feasible routes. By accessing real-time traffic lights, a model predictive control (MPC) problem is formulated, where the vehicle-lane distribution is predicted with a sub-optimization of the shortest queue and combined with the predictive scheduling. Comparisons against existing methods reveal that the proposed planner exhibits significant improvements in time efficiency and successes to avoid traffic congestion, which is available athttps://github.com/ZNianHua/DMPC-planner-for-VRP