Visual Based Control Approach for Autonomous Parking

Autonomous valet parking requires roaming the vehicle within the valet and as it arrives close enough to the target parking spot, a final maneuver is performed. In this research, such a maneuver is performed within a visual servoing framework; the relative pose of the parking spot is estimated relying on a monocular camera feed within a closed loop control system. Two control algorithms are used to solve the motion control problem, one is a cascade proportional controller with which the parking can be initiated from a defined region, another is an optimization-based control law that is tailored for L curve parking cases. Control system is developed within MATLAB and Simulink relying on synthetic image of a parking spot such that an online imagery feed is acquired. The simulation is divided into modules mainly related to perception, monocular camera model, the single-track kinematic model of the vehicle and finally the control algorithms. In addition to simulation, parking spot detection techniques of real scenarios are analyzed.