Abstract: To address the inefficiency of handling environmental constraints and kinematic constraints in trajectory planning for Ackermann-steering autonomous vehicles in complex environments,this paper proposes an efficient spatiotemporal joint trajectory planning algorithm with front-end and back-end coupling.In the front-end path planning,by integrating a topology map-guided sliding window pruning strategy with a dynamic heuristic function optimization method,the search efficiency is improved.For the back-end trajectory optimization,environmental constraints are first represented by dual safety corridors.Based on the differential flatness characteristics of the Ackermann model vehicle,kinematic constraints are expressed using the vehicle’s differentially flat variables and their higher-order derivatives as optimization variables.These variables are then analytically transformed into an unconstrained space via smooth mapping and diffeomorphic transformations.Finally,gradient optimization is performed in the unconstrained space to resolve constraint-handling inefficiencies.Experiments conducted in real-world parking environments demonstrate that the proposed method achieves 47% faster path search time compared to existing methods while maintaining comparable path quality.The trajectory optimization results show that the proposed algorithm can plan high-quality spatiotemporal trajectories in real time,verifying its effectiveness.

Key words: Trajectory planning, Spatio-temporal coupling, Differential flatness, Diffeomorphic transformation, Convex optimization