Path Planning of Automatic Parking System by a Point-Based Genetic Algorithm

Zijia Li, Fangqing Gu

Abstract: The path planning module in automatic parking systems plays a crucial role in enhancing performance and user experience. To tackle this vital aspect, we propose a path planning algorithm that utilizes a point-based genetic algorithm (POGA) in automatic parking systems. Initially, we generate some control points randomly within the parking range and then establish trajectories between each pair of control points using a cubic spline curve. In addition, we incorporate a penalty term to increase the distance between the vehicle and obstacles along the collision path. By implementing this technique, we successfully obtain the initial shortest path that ensures safety, along with the corresponding control points. Finally, POGA is proposed to conduct an optimal search for control points, in which each control points is an individual. This algorithm combines the well-established Dijkstra algorithm with a genetic algorithm, thereby improving stability and convergence. Through comprehensive comparisons with a path-based genetic algorithm, our experimental results demonstrate the superior stability and faster convergence rate exhibited by POGA across various parking scenarios.