Reinforcement Learning Improved Control Architecture for Following Vehicle in Platooning Driving
摘要
To improve the mobility and range of Platooning Driving for material transportation or regional inspection tasks in off-road environments, this paper proposed an reinforcement learning improved architecture for the speed and energy control for the following vehicle, the primary task of the architecture is to fulfill the stable and high maneuverability requirement of the following vehicle, the secondary task is to enhance the fuel economic efficiency of the engine- generator sat.For the speed control system, the Deep Deterministic Policy Gradient (DDPG) algorithm is used to control the speed which is critical for the following task. For the energy management system(EMS), a n-step improved Twin Delayed Deep Deterministic Policy Gradient (TD3-Nstep) algorithm is built to control the engine-generator set for the power supply purpose, instead of power following control method, the EMS aims at stabilizing the bus voltage of the vehicle, which can support the maneuvering characteristics of the platooning driving. Simulation experiments show that the proposed algorithm enables the platooning vehicle to have good performance in high-mobility characteristics and energy system stability.