Collision-risk-aware eco-trajectory planning of connected multi-agent vehicle platoons at signalized intersections

Eco-trajectory planning plays a critical role in alleviating urban traffic congestion and reducing vehicle energy consumption and carbon emissions. However, most existing studies focus on trajectory optimization under idealized conditions, neglecting dynamic traffic fluctuations and safety constraints. As a result, theoretically optimal trajectories often fail to cope with collision risks encountered in real-world applications. In addition, the majority of current research addresses vehicle platoons from a macroscopic perspective, lacking microscopic trajectory planning for individual vehicles within the platoon. To address these limitations, this study considers connected and automated vehicle platoons at signalized intersections and proposes a two-stage eco-trajectory planning framework. In the first stage, a terminal state prediction method based on the Lighthill-Whitham-Richards (LWR) traffic flow model is developed to capture queue evolution and shockwave propagation without requiring large-scale training data, thereby providing physically interpretable and robust boundary conditions for trajectory optimization. Based on the predicted terminal states, a dynamic programming (DP) approach is employed to generate energy-optimal trajectories for individual vehicles, enabling global energy optimization at the platoon level. In the second stage, a real-time mode-switching mechanism is introduced, which transitions from DP-based trajectory tracking to a car-following model when collision risk is detected, and smoothly recovers the reference trajectory once safety conditions are restored through a headway-feedback-based transition strategy, thereby avoiding discontinuities during mode switching. Simulation results in SUMO demonstrate that the proposed framework achieves an average energy reduction of 29.78%. The benefit is particularly evident in large platoons, where the energy-saving rate is further increased by 23.71% and 24.09% compared with adaptive cruise control (ACC) and the “1+m” strategy, respectively. While achieving energy savings, the proposed framework also ensures safe vehicle operation by effectively mitigating collision risks compared with pure DP-based trajectory tracking.