Cooperative Graph Search Using Fractal Decomposition.pdfVIP

Cooperative Graph Search Using Fractal Decomposition∗ ˜ James R. Riehl andJoaoP.Hespanha† Abstract— We present an algorithm based on hierarchical problem in continuous time and space [8], but these generally decomposition that finds close-to-optimal search paths for a apply to a very restrictive set of initial target distributions. A cooperative team of agents searching for one or more targets more practical approach is to discretize the search space and on a graph. The method partitions the graph to create a high-level problem and several lower-level problems. Since formulate the search problem on a graph. Then the search the computations on each level are identical, the lower-level path describes a sequence of regions to search along with problems can be further decomposed. In this way, the problem an amount of time that should be spent in each region. becomes fractal in nature. We use best-case and worst-case Although this discretization simplifies the problem to some instances of the decomposed problem to establish upper and extent, it is still computationally very difficult. Trummel and lower bounds on the optimal search reward, and the bounds are determined with much less computation than what is required Weisinger showed that the single-agent search problem is to solve the full problem. We show that as the number of NP-Hard even for a stationary target [13]. Eagle and Yee decomposition levels increases, the computational complexity [2] and Stewart [11] formulated the moving target problem approaches O(n) at the expense of looser boun