Vehicle platoon control in complex dynamic environments necessitates a rigorous balance between physical interpretability safety constraints and nonlinear compensation. This paper presents PERL, a Physics-Enhanced Residual Learning framework that hierarchically integrates Distributed Model Predictive Control (DMPC) with Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. To address the pathological residual action saturation. we propose a residual-magnitude-based action regularization mechanism. This mechanism penalizes large-amplitude residuals to stabilize the policy update and constrain the RL module within a safe, small-amplitude regime, ensuring effective compensation for unmodeled dynamics beyond the physical baseline’s fidelity. Additionally, a spatiotemporal smooth handover and dynamic fusion strategy significantly enhances edge-case safety. Utilizing safety-boundary constraints and dual soft-filtering, the strategy regulates control authority to mitigate violations of Rational Driving Constraints by RL policies while preventing abrupt control mutations. Experiments across four representative scenarios show that PERL converges rapidly and consistently outperforms MPC, PPO, and TD3 in safety, comfort, and platoon stability. Specifically, PERL reduces inverse time-to-collision by 20.3% relative to PPO and improves ride comfort by over 60% compared with TD3. Ultimately, the framework attains the lowest cumulative damping ratio, demonstrating superior string stability and robust generalization capability in real dynamic traffic environments.
- Article type
- Year
Open Access
Research Article
Just Accepted
Open Access
Research Article
Issue
To address the complexity of longitudinal-lateral vehicle dynamics in platoon control and the challenge of mitigating traffic oscillations, a novel model-based reinforcement learning (MBRL) method with planning capability is proposed for longitudinal-lateral control of vehicle platoons. Specifically, the integration of the environment model and policy model, combined with a sampling-based planning approach, enables planning-based control during decision-making, which significantly improves the stability and safety of vehicle platoon control. In the two-dimensional scenario, a unified state space and joint action space are designed to achieve longitudinal-lateral control, along with a multidimensional reward function that integrates both control objectives. To address the low exploration efficiency and the high proportion of ineffective exploration during the early stage of training, a prior-knowledge-guided exploration strategy is introduced. This strategy improves learning efficiency and accelerates convergence by incorporating guided actions and constraints. The training results indicate that the incorporation of prior knowledge significantly enhances training efficiency. In evaluations under speed-limit scenarios and real-data scenarios, the proposed method demonstrates superior performance in longitudinal-lateral control, platoon stability, safety, and adaptability, while maintaining high computational efficiency.
京公网安备11010802044758号