Cooperative Energy Management Method for Plug-In Hybrid Electric Vehicles Considering Interaction with Speed Planning Optimization

The fuel economy of the plug-in hybrid vehicle is controlled by the Energy Management Strategy, in autonomous driving. The coupling between the energy management and auto speed control makes the potential of cooperative Energy Management Strategy exist. This paper proposes an cooperative energy management method for in car-following scenario, considering interaction with speed planning optimization, in order to reduce fuel consumption while ensuring the safety and comfort. Within the Q-decomposition Multi-agent Independent eXtension (QMIX) reinforcement learning framework, the speed planning agent and the energy management agent are constructed separately, with their interaction constructed. The framework allows better evaluation of individual agents’ contribution to the overall reward. The proposed algorithm reduces fuel consumption by 12.3% while ensuring safety and comfort, compared to the hierarchical model predictive control (MPC) strategy.