Path Planning of Multiple Agents Through Probability Flow

A multi-agent problem arises when there are multiple intelligent agents that cannot avoid considering each other’s presence. In the context of path planning, this translates into all those situations where multiple agents have to move in the same spatial environment, such as air-traffic control, robotic warehouse management, autonomous driving, etc. In this work, we propose a fully probabilistic approach to solving multi-agent path planning by mapping the intentions of other agents within a probabilistic flow. The coordination among different agents emerges naturally and each agent can reach its own goal without generating spatio-temporal conflicts.