Reinforcement Learning-Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle

This paper presents a reinforcement learning (RL)-based energy management strategy for a hybrid electric tracked vehicle. A control-oriented model of the powertrain and vehicle dynamics is first established. According to the sample information of the experimental driving schedule, statistical characteristics at various velocities are determined by extracting the transition probability matrix of the power request. Two RL-based algorithms, namely Q-learning and Dyna algorithms, are applied to generate optimal control solutions. The two algorithms are simulated on the same driving schedule, and the simulation results are compared to clarify the merits and demerits of these algorithms. Although the Q-learning algorithm is faster (3 h) than the Dyna algorithm (7 h), its fuel consumption is 1.7% higher than that of the Dyna algorithm. Furthermore, the Dyna algorithm registers approximately the same fuel consumption as the dynamic programming-based global optimal solution. The computational cost of the Dyna algorithm is substantially lower than that of the stochastic dynamic programming.