Ecological learning-enhanced hierarchical collaborative control for fuel cell electric vehicle platoons

With growing demand for efficient, safe, and sustainable transportation in urban and intercity contexts, challenges like traffic congestion and vehicle coordination have become more prominent. Closely spaced vehicle platoons boost roadway capacity, improving traffic flow and fuel efficiency. However, realizing their full potential relies on effective coordination and control strategies. This paper proposes an ecological learning-enhanced hierarchical collaborative control (eLHCC) framework for fuel cell electric vehicle (FCEV) platoons, aiming to ensure stable car-following performance and maximize energy savings. The eLHCC comprises a higher-level controller and three lower-level controllers. The higher-level controller employs learning-enhanced deep deterministic policy gradient (LEDDPG) to exploit multi-vehicle interactions and prioritized experience replay, generating optimal reference speed datasets. The lower-level controllers utilize learning-based collaborative model predictive control (LCMPC), integrating an extreme gradient boosting (XGBoost) module to refine reference datasets. Additionally, XGBoost provides high-precision state observations, enabling accurate detection of dynamic changes and enhancing adaptability, robustness, and energy efficiency. Simulation evaluations and hardware-in-the-loop (HIL) tests confirm the proposed strategy's effectiveness in improving safety, stability, and economic efficiency under diverse conditions, demonstrating its potential for advancing energy management and coordination in multi-energy-source vehicle platoons.