Abstract
We present a high-speed planning method with compact precomputed search trees using a new pruning method and evaluate the effectiveness and the efficiency of our precomputation planning. We define two obstacle rates which express a dispersion in a real indoor environment, and we introduce the limitation of precomputation planning using these obstacle rates. Its speed is faster than an A* planner in maps in which the obstacle rate is the same as indoor environments. Precomputed search trees are one way of reducing planning time; however, there is a time-memory trade off. Our precomputed search tree (PCS) is built with pruning based on a rule of constant memory, the maximum size pruning method (MSP) which is a preset ratio of pruning. We apply the node selection strategy (NSS) to MSP. We extend the outer edge of the tree and enhance the path reachability. Additionally, the alternate branch backtracking (ABBT) enhances a success rate in crowded environments. In maps less than 30% obstacle rates on a map, the runtime of precomputation planning is more than one order of magnitude faster than the planning without precomputed search trees. Our precomputed tree finds an optimal path in maps with 25% obstacle rates. Then our precomputation planning speedily produces the optimal path in indoor environments.
