Path planning for autonomous vehicles in complicated environments

When the autonomous vehicle is running in complex unstructured road and the reference path to guide the vehicle could not be obtained directly, at this time, only sparse task list point left as the reference information. To generate a safe and smooth path in this situation for autonomous vehicle to run, we propose an improved heuristic graph search path planning algorithm to solve the problem by considering both the obstacle constraint and the vehicle model. To ensure the planning result is executable, kinematic constraints of the vehicle is considered when designing the motion primitives; for the efficiency and safety, obstacle constraint and the vehicle motion were taken into account when designing the cost and heuristic functions; in order to ensure that the vehicle could accurately reach the target pose, Reeds-Shepp curves are used to connect the target pose, and numeric non-linear optimization method was used to improve the smoothness of the plan result. Experimental results in the autonomous vehicle show that the methods in this paper is effective.