Planning two-dimensional trajectories for modular bus enroute docking

Modular buses (MBs), which can physically dock and separate, offer enhanced flexibility and potential cost savings in urban transportation. Despite advances in scheduling, trajectory planning for the docking process of MBs is less developed. This paper addresses the two-dimensional trajectory planning for MB docking. We introduce a hierarchical docking planning model based on Nonlinear Model Predictive Control (NMPC). The upper-level model optimizes docking time and speed, while the lower-level dynamically updates trajectories. Our models integrate Frenet and Cartesian coordinates with a precise obstacle avoidance model to ensure safety and smoothness under diverse traffic conditions. We employ segmented Lagrange interpolation for discretizing the continuous NMPC model, enhancing planning accuracy with fewer points and improving solving efficiency. Additionally, a multi-task network adaptively adjusts discretization orders based on environmental data. Extensive testing demonstrates our method's superior accuracy and efficiency in real-time performance, offering marked improvements in safety and operational smoothness compared to existing approaches.