A path planning algorithm for parking an autonomous vehicle.
Recently, with the development of self-driving, autonomous vehicles and automated warehouses there have been a growing interest in the development of autonomous path planning as well as parking algorithms. In this study, a solution for path planning in autonomous parking for 4-wheel-AGV(Automated Guided Vehicle) is presented. In the theoretic part of the study 3 path planning algorithms are exposed such as: Dubbin’s path, RRT (Rapidly-exploring random tree) algorithm and an arc path planning algorithm. In the practical part of the study the arc path parking algorithm is implemented and tasted.
The main focus is given to algorithm testing and presentation of obtained test results in simulation environment. The path is circle arc that connects the vehicle initial position and goal position of the vehicle inside the parking space. The algorithm is tested in a simulation environment where parking space and dimension of the vehicle are known. At the beginning from the initial position of the vehicle to the final position the algorithm find more than one potential final circle that can be used to defending the final arc path. But the circle that is picked as final path has to have radius larger than the turning radius of the vehicle and the tangent point between the potential final circle and circles with turning radius of the car have to be above X axis of the local coordinate system.
The proposed path planning algorithm provides a practical method of path planning without expensive computation and it was proved that if find a final path from any initial position. In the study it was proved that the length of the final path in changing depending on this parameters: position in orientation of the vehicle, steering angle and parking forward/backward. In the study 4 different optimizations to find the shortest path were done so the optimal path can be found. First one is optimization of the positions and orientations of the vehicle where we know the steering angle thata. With this optimization the best area where the vehicle should start with the parking procedure and path planning and the best orientation of the vehicle are found. Third is optimization of the steering angle theta, where we have the initial starting position and orientation of the vehicle. With this optimization we get what is the best steering angle of the vehicle so we get the optimal path length. And the last optimization is with parking forward/backward. With the optimization it was shown that the length of the path is shortest when vehicle park forward if it is turn with the front side of it to the parking spot. And if it park backward it is better if it is turned with the back towards the parking spot. Also while doing all the optimizations it was shown that the shortest path is found if the tangent point between final circle and circle with turning radius is same as the position point of the vehicle.
The good side of the algorithm is that it always find a path, without expensive computations. But the path might not be always optimal, that is why it is good to know the optimization parameter that is important. If it is searching for optimal path meaning the shortest path, or the path with defined number of maneuvers or other optimization.
The further work on the algorithm would be implementing and testing in on a real example in real environment.
|
