Reinforcement learning-based real-time energy management for a hybrid tracked vehicle

Abstract

To realize the optimal energy allocation between the engine-generator and battery of a hybrid tracked vehicle (HTV), a reinforcement learning-based real-time energy-management strategy was proposed. A systematic control-oriented model for the HTV was built and validated through the test bench, including the battery pack, the engine-generator set (EGS), and the power request. To use effectively the statistical information of power request online, a Markov chain-based real-time power request recursive algorithm for learning transition probabilities was derived and validated. The Kullback–Leibler (KL) divergence rate was adopted to determine when the transition probability matrix and the optimal control strategy update in real time. Reinforcement learning (RL) was applied to compare quantitatively the effects of different forgetting factors and KL divergence rates on reducing fuel consumption. RL has also been used to optimize the control strategy for HTV, compared to preliminary and dynamic programming-based control strategies. The real-time and robust performance of the proposed online energy management strategy was verified under two driving schedules collected in the field test. The simulation results indicate the proposed RL-based energy management strategy can significantly improve fuel efficiency and can be applied in real time.