A new approach to cooperative pathfinding (Short Paper)

In the multi-agent pathfinding problem, groups of agents need to plan paths between their respective start and goal locations in a given environment, usually a two-dimensional map. Existing approaches to this problem include using static or dynamic information to help coordination. However, the resulting behaviour is not always desirable, in that too much information is hand-coded into the problem, agents take paths which look unintelligent, or because the agents collide and must re-plan frequently. We present a distributed approach in which agents share information about the direction in which they traveled when passing through each location. This information is then used to encourage agents passing through the same location to travel in the same direction as previous agents. In addition to this new approach, we present performance metrics for multi-agent path planning as well as experimental results for the new approach. These results indicate that the number of collisions between agents is reduced and that the visual fidelity is improved.