Simulation Study of Processes in Electric Vehicles under Braking Control Based on Reinforcement Learning

Uncertainty and non-linearity in the braking process of electric vehicles prevent the use of classical control methods. Given the presence of many algorithms, each of which acts differently in various braking conditions, this study proposes an intelligent braking methodology that can adapt to changing modes and environmental characteristics. Once the need in braking is detected, the required parameters of the vehicle and the road are forwarded into the neural network generated its action with the help of deep reinforcement learning. A system presented makes a decision regarding the braking torque strength and its sharing between electrical and friction brakes for every driving situation. The simulation was carried out to test the effectiveness of the proposed approach which results show that the offered method ensures high-quality deceleration with good energy recovery without skidding and, thereby increases the braking efficiency.