Abstract:
Multi-agent coordination and routing is a complex problem and has a wide range of applications in areas from vehicle fleet coordination to autonomous mapping. Whereas traditional methods are not designed for realistic environments such as sparse connectivity and unknown traffics and are often slow in runtime; in this paper, we propose a graph neural network based model that is able to perform multiagent routing in a sparsely connected graph with dynamically changing traffic conditions, outperforming existing methods.
Our learned communication module in the proposed model enables the agents to coordinate online and adapt to changes to their environment. We also show that our model trained with only two agents on graphs with a maximum of twenty-five nodes can easily generalize to five agents with a hundred nodes.
