Successive Convex Approximation-Based Fast Collision-Free Motion Planning Framework for Autonomous Parking

Motion planning is a challenging problem in autonomous driving and robotic control, especially planning a collision-free trajectory for autonomous parking in narrow environments. Optimization-based motion planning is a popular method, since it may provide an optimal trajectory. Unfortunately, to avoid obstacle, it is always a non-convex optimization programming in real world application. Hence, it may be a high computational complexity and time-consuming task. To address this issue, this paper presents a successive convex approximation-based fast collision-free motion planning framework to address this issue. The key idea is to convert non-convex collision avoidance constraints into smooth convex constraints and use the successive convex approximation method to improve the computational efficiency. Furthermore, we incorporate the Hybrid A* algorithm to provide good initial guesses and use the Armijo step-size rule to prevent overshooting optimal solutions. The simulation results show that, compared to traditional benchmark methods, the proposed approach improves computational efficiency by an average of 83.6% in scenarios such as vertical, while also generating smoother trajectories. The method's efficiency and robustness make it more suitable for real-time deployment on low-computation platforms.