Trajectory planning for autonomous semi-trailer vehicles based on EDT constraints

Real-time trajectory planning for autonomous semi-trailer vehicles is challenging because tractor–trailer coupling and trailer tail-swing make it difficult to simultaneously ensure kinematic feasibility, obstacle clearance, motion smoothness, and computational efficiency. This paper proposes a trajectory-planning method with Euclidean distance transform (EDT) constraints that combines motion-primitive graph search with online nonlinear model predictive control (NMPC). The graph-search stage generates a collision-free initial trajectory satisfying the tractor–trailer kinematic constraints, while the Euclidean distance transform converts discrete obstacle information into continuous clearance information. The EDT distance field is incorporated into NMPC as both a soft obstacle-avoidance cost and a hard minimum-distance constraint. Simulation results in structured and unstructured environments show that the proposed method provides larger safety margins and smoother trajectories than the timed elastic band (TEB) planner, while reducing fluctuations in the tractor steering angle and trailer articulation angle. The average planning times are 33 and 37 ms, respectively. Real-vehicle experiments further verify the practical executability of the planned trajectories.