Energy-efficient robust control of vehicle platoons under cut-in disturbances: Integrating temporal-aware policy and barrier-constrained search

With the increasing demand for efficient and safe transportation, vehicle platooning has emerged as a pivotal solution to enhance roadway capacity and energy economy. However, realizing its full potential hinges on overcoming unmodeled dynamic uncertainties and external stochastic disturbances, particularly non-cooperative cut-in maneuvers. This paper proposes a novel hierarchical cooperative control architecture that synergistically integrates a low-level cooperative execution layer with a high-level adaptive reference layer, aiming to ensure robust car-following performance and maximize energy savings. In the high-level module, the adaptive spacing policy based on Barrier-Constrained Stochastic Search (BCSS) employs a decoupled optimization architecture, leveraging barrier constraint theory to formally establish a verifiable safe-set via a novel predictive dynamic barrier function, while simultaneously applying stochastic search to identify the optimal reference that aligns with comfort and economic tendencies. Complementing this, the low-level Temporal-Aware Recurrent Deterministic Policy (TRDP) controller architecturally enhances the platoon collaboration by integrating Long Short-Term Memory (LSTM) cores, which enables the agent to effectively capture complex temporal dependencies and reconstruct the underlying state evolution, thereby generating a proactive cooperative control policy that ensures string stability and economic efficiency. Comprehensive simulations and scaled vehicle platform experiments demonstrate that the proposed architecture significantly enhances platoon robustness and energy-saving performance while maintaining superior disturbance suppression performance.