Although several (semi-) automatic parking systems has been proposed throughout the years, no solutions have been found so far using on-the-market sensors for performing closed-loop maneuvers with an optimal trajectory based on a real-time assessment of the parking spot. In the light of these observations, we investigated how Local Motion Sensors (LMSs) providing 1-D optical flow (OF) measurements could be used to improve automatic parking maneuvers.
For this purpose, we propose a low computational-cost method to detect and track a parking spot in real time using 1-D OF measurements around the vehicle as well as the vehicle’s longitudinal velocity and steering angle. The results of a highly simplified 2-D parking simulation show that up to a certain noise level, the simulated vehicle can detect and track the parking spot in real time.
The preliminary results with two 6-pixel LMSs on a vehicle show that the LMSs can provide an average refresh frequency from 2 to 4 times higher than standard ultrasonic sensors and cameras, and can therefore represent a good solution for new automatic parking systems as well as other automotive applications.