Deep reinforcement learning based active safety control for distributed drive electric vehicles

Distributed drive electric vehicles are regarded as the promising transportation due to advanced power flow architecture. Optimizing yaw motion enhance vehicle safety is a challenging job. Besides, nonlinear features in affect control accuracy of controllers. To this end, deep reinforcement learning (DRL) based direct moment (DYC) strategy put forward here. Vehicle dynamics can be approximated with DRL algorithm, which reduces complex solving process. Concretely, DYC problem formulated Markov Decision Process observed signals and external incorporated state action sets. Thereupon, actor-critic network exhibited approximate action-value function policy for better performance. Furthermore, guarantee continuous solution moment, deterministic gradient algorithm employed, target online parameters simultaneously trained maintain process stability. The proposed verified using Carsim/Simulink platform under typical lane change manoeuvres. Numerical test results demonstrate that outperforms linear approaches on taking full advantage understeer enhancing lateral stability, especially critical steering