Real Time Voronoi-like Path Planning Using Flow Field and A*

Abstract

In this paper we introduce a path planning approach for robotic and related applications to reach a target within a known environment. We improve the flow fields based path planning approach by avoiding typical obstacle hugging behavior. Flow field path planning defines the potential at each node of an environment which allows a large number of agents to traverse the environment while avoiding the task of calculating paths for each individual agent. This approach maintains the same benefit of typical flow field path planning with the addition of avoiding paths that are too close to obstacles. This method can be referred to as Valley Field Path Planning. In the general method, this algorithm takes two major steps. Firstly, the obstacle potential field that results in a low potential in regions farthest from the obstacles. The second step of the algorithm is to find potential provided that the goal node corresponds to a higher potential at nodes farther. The weighted sum of these two potential fields results in a flow field that avoids obstacle hugging. To avoid the local minima problem with this approach, $A^{\ast}$ path planning can been introduced with the benefit of a configurable deterministic path at the cost of calculating the path on a per agent basis. The results show that this combination of an obstacle potential field and the A* search algorithm results in an efficient and inexpensive way to generate a configurable path.