Sampling based vehicle motion planning for autonomous valet parking with moving obstacles

Abstract

This paper describes a motion planning algorithm for unstructured dynamic environments with motion prediction for moving obstacles. The proposed algorithm is composed of the four steps: 1) target motion prediction; 2) drivable area decision 3) local path planning and 4) vehicle control. The target motion prediction is crucial parts for realizing autonomous valet parking system because many vehicles which search available parking lot exist simultaneously. To predict future motion of target, the intention of the target should be inferred first. Interacting multiple model (IMM) filter using two models has been used to infer the intention of the target. Based on the inferred intention, most appropriate model's results are used as a predicted trajectory of the target vehicle. After that, the drivable area is decided to avoid collision with static obstacles and moving targets using potential filed approach to assessment the risk. In this stage, pre-defined parking lot map which contains boundary of the parking lots and waypoints is used to define initial guess of the drivable area. Inside the drivable area, rapidly-exploring random tree (RRT) generate the desired local path while guaranteeing the real-time performance in dynamic environments. Finally, path tracking controller and speed controller calculate desired steering wheel and longitudinal acceleration input. The proposed motion planning algorithm is validated via MATLAB based computer simulation. Simulation results demonstrate the ability of the proposed motion planning algorithm for unconstructed dynamic environments to plan collision-free path which is appropriate in parking lot situations.