Abstract: To improve safety, parking efficiency, smoothness and riding comfort of parallel parking of driverless vehicles at roadside, the characteristics of parallel parking conditions were analyzed, and the path planning was carried out by quintic polynomial curve. The mathematical model of path planning curve parameters optimization design was established with the abscissa of parking starting position as design variable, and six equality constraints were established based on parking boundary conditions. Three inequality constraints were established based on the requirements of curvature, collision avoidance and reducing parking space, and the weighted sum of the maximum curvature and the absolute value of abscissa of parking starting position was set as objective function. By the particle swarm optimization algorithm, the abscissa of the parking starting position was obtained, and six polynomial undetermined coefficients were calculated. The optimized smooth and continuous curvature planning path curve expression was obtained. Parking path tracking was realized based on the pure pursuit algorithm. The results show that the vehicle can track the optimized path precisely without collision with obstacles, which meets the safety requirements. The steering wheel angle changes continuously to meet the requirements of smoothness and ride comfort.