Reinforcement Learning based Energy Management system for Hybrid Electric Vehicles

Abstract

Hybrid Electric Vehicles (HEV) (Self-charging Hybrid Vehicles) have been exemplified as the timely, practical, balanced solution for fuel-efficiency and eco-friendly environment. Considering the design perspective of the powertrain, the different strategies employed for Energy Management have depicted vast improvements in the energy efficiency of the hybrid powertrain models. With companies like deep-mind promoting Reinforcement Learning (RL) with the human-level efficiency to task multi-games, RL-based algorithms have been arduously researched to develop Energy Management System (EMS) owing to the capability to self-learn interacting with any complex environment by the return in rewards for each action. The recent research proved this to be on a larger efficient scale compared to the predefined algorithms. This article reviews the different powertrain models and their working along with some basic RL algorithms for any alterations from the pre-existing models to improve the efficiency and develop the easiness of the management system. Ample importance is given to the classification of EMS, and the primary EMS created using RL in this review. A model for the EMS is also discussed along with the results. The future scope of work is also discussed.