Learning Time Reduction Using Warm-Start Methods for a Reinforcement Learning-Based Supervisory Control in Hybrid Electric Vehicle Applications

Reinforcement Learning (RL) is widely utilized in the field of robotics, and as such, it gradually being implemented Hybrid Electric Vehicle (HEV) supervisory control. Even though RL exhibits excellent performance terms fuel consumption minimization simulation, large learning iteration number needs a long time, making hardly applicable real-world vehicles. In addition, initial phases much worse than baseline controls. This study aims to reduce iterations Q-learning HEV application improve utilizing warm start methods. Different from previous studies, which initiated with zero or random Q values, this initiates different controls (i.e., Equivalent Consumption Minimization Strategy control heuristic control), detailed analysis given. The results show that proposed requires 68.8% fewer cold Q-learning. trained validated two driving cycles, 10-16% MPG improvement when compared Furthermore, real-time feasibility analyzed, guidance vehicle implementation provided. can be used facilitate deployment applications.