Path Optimization-Based Online Data-Driven Obstacle Avoidance Trajectory Planning for Intelligent Vehicles

Fast obstacle avoidance trajectory planning has significant application value for intelligent vehicles. This work proposes an online data-driven method for trajectory planning, which allows for low computational complexity by using off-line computations to speed up the process in real-time. Firstly, the optimal control problem (OCP) is formulated to avoid collisions with static obstacles and road boundaries. Subsequently, the Gaussian pseudo-spectral optimization technique is employed to solve the OCP by utilizing a collocation strategy with different constraint configurations and then generate the optimized path trajectories. Next, an obstacle avoidance optimization database is established by solving obstacle avoidance OCPs off-line. Accordingly, an optimization data-driven trajectory planning strategy is designed to achieve online planning and the fuzzy logic is proposed to generate feasible obstacle avoidance trajectories. Simulation tests demonstrate that the proposed method effectively reduces trajectory planning time by over 98% while generating similar trajectories, as evidenced by a comparison with the Gaussian pseudo-spectral method and Bezier curve method. Ultimately, real vehicle tests verify the performance of the proposed method for online obstacle avoidance.