Coordination Control of Differential Assisted Steering and Stability for Distributed Drive Electric Vehicles Based on Twin Delayed Deep Deterministic Policy Gradient

Differential Drive Assist Steering (DDAS) improves vehicle maneuverability by applying a differential torque between a vehicles's the left and right wheels. However, since DDAS and Vehicle Stability Control utilize the same actuator set, dynamic conflicts may arise, leading to degraded performance and potential instability. To resolve this issue, we propose a novel coordination control framework based on deep reinforcement learning, integrating a dynamically weighted reward function with a phase-plane-based adaptive weighting strategy. The proposed method optimizes the trade-off between steering assistance and stability by adjusting control inputs in real time according to the vehicle's operational state. For validation purposes, we conducted a co-simulation study in a Carsim/Simulink environment to test the effectiveness of the method, demonstrating significant reductions in steering effort and enhanced handling stability without compromising safety. The results indicate that the proposed method outperforms conventional control strategies in mitigating actuator interference while ensuring robust performance under diverse driving conditions.