Efficient Path Planning and Tracking Control of Autonomous Vehicles in Park Scenarios Based on Sparrow Search and Stepwise Prediction

Electrification and intelligence dominate the developing trend of automobiles. Except the tremendous progress in electrification, autonomous driving is widely investigated with the advancement of perception and communication technologies. While, proper path planning and rigorous tracking control in autonomous driving remain challenging, especially in park scenarios. To cope with it, an efficient long-short term path planning algorithm is firstly proposed to compensate the lack of global driving information in park scenarios, and a fifth-order polynomial and an improved artificial potential field method are incorporated to devise the candidate trajectory clusters for obstacle avoidance. Secondly, the collision detection of candidate trajectories is attained with the support of a rectangular vehicle model, and the dangerous trajectories of potential collisions in the candidate clusters are screened out. Then, three evaluation indicators considering safety, comfort and lane change efficiency are weighed to establish the multi-performance objective function, and the sparrow search algorithm is exploited to determine the optimal obstacle avoidance trajectory. Finally, the stepwise prediction with multiple prediction models is exploited to achieve precise trajectory. The simulation and experiment results manifest that the proposed method enables to achieve precise path planning and tracking, and ensures satisfactory safety, comfort, and real-time performance in park scenes.