Adaptive Reinforcement Learning Framework for Optimized Energy Management in Hybrid Electric Vehicles Using Synthetic Driving Environments and Open-Source Simulation Models

Conventional heuristic approaches are sometimes used to model energy management strategies for HEVs, especially for Rule-Based car models, but RL is the solution that presents a shift in strategy. In this research, RL is used to solve the problem of M/M objectives in the field of HEVs by optimizing fuel efficiency and emissions. Based on the realistic synthetic driving cycles that mimic various realistic driving scenarios, and the open-source toolbox OpenPowertrainLib, the research develops a reproducible and computationally saving approach. I have suggested that the powertrain energy distribution should be trained through reinforcement learning (RL), PPO, a policy-gradient-based RL algorithm should be used to propose an agent capable for adaptive control. When it comes to implement the maneuverer the dynamics of the hybrid vehicle as well as the constraints of the powertrain are modelled within the OpenPowertrainLib’, which provides credible and reconstructive looks at the maneuverer under simulation. The simulations show a remarkable increase in energy efficiency and emissions decrease in comparison with conventional heuristic methods. This work helps towards sustainable transportation by demonstrating the promise of RL-driven intelligent energy management systems of HEVs.