Gradient-Optimized B-spline Trajectory Planning for Unmanned Platforms

Local trajectory planning plays a pivotal role in ensuring the core functionality of autonomous driving systems. This paper presents a fast and lightweight B-spline curve optimizer for local trajectory planning in autonomous driving systems, incorporating three key metrics: safety, smoothness, and efficiency. The algorithm utilizes collision and collision-free trajectory gradient information obtained through A* search for efficient collision distance estimation, thereby generating safety metrics during trajectory optimization. The unconstrained optimization problem is addressed using L-BFGS, significantly improving solution efficiency. The algorithm’s effectiveness and real-time performance are validated through simulations and practical experiments.