Planning to find an unpredictable evader using rapidly-exploring random tree

We propose an efficient algorithm for a single pursuer searching for an unpredictable static evader in a 2D environment. The algorithm uses a newly emerging data structure, which is the rapidly-exploring random tree. This data structure is constructed randomly but evenly distributed to cover the whole workspace. We propose the region-based sampling strategy to insure the full uniform coverage of the workspace. During construction of the tree, we keep checking for the evader, which are the stopping criteria of the algorithm. However, if the algorithm reaches the anticipated coverage percentile without detecting the evader, then the algorithm reports failure (i.e., evader is blocked). A lazy collision detection approach is used to resolve any collision with the obstacles during runtime. Simulation results demonstrate the validity of the algorithm.