Vision-based autonomous indoor valet parking system

Abstract

This paper briefly introduces our on-going research on in-door autonomous valet parking (AVP) using only camera vision sensors. The AVP consists of autonomous driving and parking. Autonomous driving utilizes waypoints given by an array of ceiling LEDs to develop a driving path for ultimate waypoint tracking. The LEDs are detected by a top-front-view camera mounted on the vehicles windshield. Autonomous parking utilizes parking stripes to detect parking vehicles pose with respect to the target parking spot. The parking spot is detected by a rear-view camera mounted on the vehicles rear window. A clothoid parking path is generated with surround obstacles being considered to complete autonomous parking. The AVP research is already showing some prospective results. Some experimental results of waypoint and parking spot detections are presented, together with some simulation results of path planning and control.