Learning based predictive control of truck platoons with speed planning

Connectivity and autonomy stand out as two highly promising technologies for fuel efficiency in the realm of automated transportation. This paper addresses the issue of fuel consumption in autonomous truck platoons and introduces a hierarchical framework. Within the speed planning layer, a fuel consumption model for the leading truck is formulated to compute velocity profile. Simultaneously, a distributed learning-model predictive control (DL-MPC) method is employed to ensure the cohesive movement of a truck platoon in a predefined formation. The co-simulation platform, leveraging Trucksim and Simulink, provides a comprehensive environment for assessing and validating the proposed control strategy. Comparison experiments validate the effectiveness of the proposed method. Distributed model predictive control (DMPC) and two-delay deep deterministic policy gradient (TD3) algorithms are conducted to illustrate the advantages of the controller in terms of control performance and computation time. The speed planning method outperforms constant speed, experience-based speed setting, and reinforcement learning-based speed planning in terms of fuel efficiency. Furthermore, the energy-saving effect of the proposed strategy is verified from the perspective of engine fuel characteristics. Therefore, the framework proposed in this paper ensures excellent control performance and fuel economy of truck platoon while reducing the computation time of the controller.