Real-time path planning and following for nonholonomic unmanned ground vehicles

Abstract

This paper describes a computationally cost-effective path planning method by combining a hybrid A∗ path planner with potential fields. The proposed real-time path planner is capable of finding the optimal, collision-free path for a non-holonomic unmanned ground vehicle (UGV) in an unstructured environment. First, a hybrid A∗ path planner is designed to find the optimal path through connecting the current position of the UGV to the target in real-time while avoiding any obstacles in the vicinity of UGV. The advantages of the developed path planner are that, by using the potential field techniques and by excluding the nodes surrounding every obstacles, it significantly reduces the search space of the traditional A∗ approach; it is also capable of distinguishing different types of obstacles by giving them distinct priorities based on their natures and safety concerns. Such an approach is essential to guarantee a safe navigation in the environment where humans are in close contact with autonomous vehicles. Then, with consideration of the kinematic constraints of the UGV, a smooth and drivable geometric path is generated. Finally, extensive practical experiments are conducted in a dynamic environment to verify the effectiveness of the proposed path planning methodology.