Multi-agent coordinated control framework for longitudinal–vertical dynamics in electric vehicles with regenerative braking

To tackle the challenge of balancing regenerative braking and ride comfort, this study presents a control framework based on the multi-agent Munchausen Prioritized Experience Soft Actor-Critic (MA-MPE-SAC) algorithm, which tightly couples braking and suspension systems to enable intelligent, dynamic coordination of energy recovery and vertical comfort in electric vehicles. The framework employs a centralized training and decentralized execution (CTDE) paradigm, where global state and joint action information are leveraged during training to enhance policy stability and convergence. Furthermore, techniques such as munchausen reward shaping, prioritized experience replay (PER), and emphasizing recent experience (ERE) are incorporated to accelerate convergence and enhance adaptability under parameter uncertainty. Under variable braking intensity conditions, the proposed method outperforms traditional decoupled RB-H2/H∞ control by improving energy recovery efficiency by 21.54%, reducing body acceleration by 14.42%, and lowering pitch angular acceleration by 30.69%. Compared to the Multi-Agent Proximal Policy Optimization (MA-PPO) algorithm, the proposed strategy demonstrates superior performance in terms of convergence, control effectiveness, and generalization capability. In addition, ablation studies confirm that the cooperative multi-agent control outperforms its single-agent counterpart with better coordination between longitudinal and vertical objectives. Real-time feasibility of the proposed algorithm is further validated through hardware-in-the-loop (HIL) experiments. These results highlight the effectiveness and engineering potential of the proposed approach for multi-objective coordination in regenerative braking and ride comfort control of intelligent electric vehicles.