MAPF-LNS2: Fast Repairing for Multi-Agent Path Finding via Large Neighborhood Search

Multi-Agent Path Finding (MAPF) is the problem of planning collision-free paths for multiple agents in a shared environment. In this paper, we propose novel algorithm MAPF-LNS2 based on large neighborhood search solving MAPF efficiently. Starting from set that contain collisions, repeatedly selects subset colliding and replans their to reduce number collisions until become collision-free. We compare against variety state-of-the-art algorithms, including Prioritized Planning with random restarts, EECBS, PPS, show runs significantly faster than them while still providing near-optimal solutions most cases. solves 80% random-scenario instances largest benchmark suite runtime limit just 5 minutes, which, our knowledge, has not been achieved by any existing algorithms.