Multi-Agent Path Finding with Priority for Cooperative Automated Valet Parking

Shortage of parking lots has been a serious problem due to the rapid population growth in urban areas. Cooperative Automated Valet Parking (Co-AVP) is a promising approach to mitigate the problem. Co-AVP system would realize high density and efficient parking by automatically navigating vehicles to vacant parking spaces. Cooperative path planning has been extensively studied as Multi-Agent Path Finding (MAPF). In Co-AVP, vehicles would be often prioritized. For instance, vehicles would be able to spend more time to park in vacant spaces than moving from parking spaces to waiting drivers. Previous MAPF settings, however, cannot treat such priorities. In this paper, we formulate MAPF considering each agent's priority. We also develop the optimal method based on Conflict-Based Search (named CBS-Pri) and the heuristic method based on Cooperative A* (named CA*-Pri). We verified that higher prioritized vehicles tended to preferentially arrive at waiting drivers by our methods in numerical experiments. We also found that CBS-Pri always searched better solutions than or equal to CA*-Pri but requires more computational time.