基于凸优化的自动泊车轨迹规划方法

A general parking path planning method was proposed to address the limitation that traditional Hybrid A* often failed to strictly satisfy vehicle kinematic constraints in parallel，vertical，and oblique parking scenarios. The parking space constraint was formulated as a convex optimization problem via the Lagrange dual function，enabling the computation of a globally optimal path under kinematic feasibility. To overcome the low efficiency of nonlinear programming，a warm-start strategy was adopted：the initial path was generated using Fault-Tolerant Hybrid A*，the reference velocity was derived from trajectory curvature，and the dual variables were initialized from the constraints. The final solution was obtained by solving the transformed nonlinear equations. Simulation studies on three parking scenarios verify the versa⁃ tility and optimality of the proposed method.