A fast analytical local path planning method with applications in parking scenarios

In this paper an analytical and geometry-based path planning method is proposed that provides non-holonomic and collision free paths for parking maneuvers. The proposed method is straightforward in the sense that it does not require any sampling or exploring of the configuration space, or relies on any optimization process. Furthermore, the calculation is fast enough to be used in real-time applications for guiding the vehicle in parking scenarios. For any pair of start-goal configurations, the planner presents an infinite set of solutions. These paths have minimum number of driving direction changes as this can be more important for a driver than the path length. They may have any curvature lower than or equal to the maximum specified curvature. Simulations show the superiority of the method in local path planning over conventional methods such as hybrid A* and optimization based approaches in terms of computation time and path quality attributes such as the number of motion primitives and the number of direction changes. To compose a complete motion planning system, the proposed local planner is applied along with hybrid A* as global planner and a path tracking controller. The performance of the motion planning system is evaluated by simulating a valet parking scenario. MATLAB/Simulink is used to simulate the system, and a comprehensive vehicle model in CARSIM is used to represent the vehicle. Simulation results shows the capability of the proposed motion planner to implement automated valet parking tasks while respecting the physical limitations of the vehicle subsystems and driving comfort requirements.