Efficient Global Trajectory Planning for AGVs Based on Convex-Concave-Decomposition

The non-convexity of constraints has long been a challenging issue in trajectory optimization problems. In this work, we propose a novel convex optimization framework for trajectory planning of autonomous ground vehicles (AGVs) that explicitly accounts for the vehicle's geometric dimensions and convex obstacles. To obtain a conservative yet effective approximation of obstacle constraints, the convex-concave decomposition (CCD) technique is employed to handle bilinear terms and equality constraints. Meanwhile, a constraint relaxation strategy is introduced to improve the numerical stability and convergence behavior of the sequential convex programming (SCP). Furthermore, a key obstacle extraction strategy is adopted to filter irrelevant obstacles and improve computational efficiency. Finally, simulations and comparisons demonstrate the effectiveness and competitive performance of the proposed framework.