Actual-Vehicle Verification of Automatic Parking in Narrow Spaces Considering Precise Collision Avoidance Constraints for Rectangular Objects

Automated parking technology is becoming increasingly popular to assist non-skilled drivers with parking. However, these systems face challenges when utilized in narrow spaces where precise collision avoidance is required. Most conventional studies approximate the controlled vehicles as circles, ellipsoids, or other shapes, which makes generating paths difficult in narrow spaces due to excess margin. To solve this problem, a new approach is proposed that uses non-linear transformations to describe collision avoidance constraints among rectangles on a manifold within the model predictive control (MPC) framework. This approach reduces the computational cost of optimization calculations compared to some conventional shape-aware collision avoidance methods. This study validated the usefulness of the proposed method by conducting tests with an actual vehicle. The results confirmed that our method enables automated parking in very narrow spaces in real-world application.