Dynamic scheduling of autonomous buses with platooning

Autonomous buses can well address issues such as driver shortages, fatigue-related driving, unsafe driving behaviors, and low punctuality in operations. With capabilities for cooperative platooning and speed guidance, autonomous buses can adapt to varying passenger demands. Nevertheless, the operator faces the dual challenges of multi-task planning and immediacy requirements. In response to these challenges, this study investigates the dynamic scheduling of autonomous buses with platooning under the demand-responsive mode. The problem is to determine supply–demand matching, departure times, speed adjustment, holding time, stop-skipping, and coupling and decoupling strategies. A vehicle energy consumption model based on vehicle-specific power is developed to assess energy use under different platoon configurations. We formulate the problem as a Markov Decision Process (MDP) model under a rolling horizon framework, and introduce a periodic state update mechanism to account for system state changes in multiple strategy execution scenarios. We propose pruning strategies to reduce the decision space. We adopt approximate dynamic programming (ADP) to solve the problem. To enhance the convergence efficiency of the ADP algorithm, we propose a multi-step look-ahead strategy based on dynamic programming (DP) to avoid the short-sighted exploitation issue of traditional ADP algorithms that rely on a single-step forward-looking strategy. We conduct the case study of the Guangzhou Tower route in the demonstration application project of autonomous buses in Guangzhou. Results show that the DP-based multi-step look-ahead strategy significantly improves the solution accuracy of the ADP algorithm. Compared to the operation without platooning, the operation with platooning can improve vehicle utilization by increasing the load factor of the fleet, while improving the service quality concerning the response rate, lateness rate and late time. The benefit of platooning is higher under large passenger demands.