Model Predictive Path Planning for Autonomous Parking Based on Projected C-Space

This paper proposes a model predictive path planning method with collision avoidance in narrow parking space. Objects are expressed as rectangle collision area, and obstacle avoidance constraints are defined by inequalities in a projected C-space which projects the rectangle collision area into the circle. Switching of driving forward and backward is planned automatically in the proposed method without a limitation on the switching. Numerical simulations were carried out to validate the performance with strict obstacle avoidance in a narrow environment assuming the reverse parking and the parallel parking tasks. The Achievability of the parking task and the computation time were compared between the proposed method and the conventional method with the collision avoidance constraints derived from Farkas' lemma.