FDA* : a focused single‐query grid based path planning algorithm

Square grid representations of the state‐space are a commonly used tool in path planning. With applications in a variety of disciplines, including robotics, computational biology, game development, and beyond. However, in large‐scale and/or high dimensional environments the creation and manipulation of such structures become too expensive, especially in applications when an accurate representation is needed. In this paper, we present a method for reducing the cost of single‐query grid‐based path planning, by focusing the search to a smaller subset, that contains the optimal solution. This subset is represented by a hyperrectangle, the location, and dimensions of which are calculated departing from an initial feasible path found by a fast search using the RRT* algorithm. We also present an implementation of this focused discretization method called FDA*, a resolution optimal algorithm, where the A* algorithm is employed in searching the resulting graph for an optimal solution. We also demonstrate through simulation results, that the FDA* algorithm uses less memory and has a shorter run‐time compared to the classic A* and thus other graph‐based planning algorithms, and at the same time, the resulting path cost is less than that of regular RRT based algorithms.