Guidance Channel Constraints and Congestion Communication for Lifelong Multi-Agent Path Finding
Yu Wang, Xuxiu Zhang, Xiaoxin Li · Advances in transdisciplinary engineering · 2025
Lifelong Multi-Agent Path Finding (LMAPF) is a variant of Multi-Agent Path Finding (MAPF) that moves multiple agents from a starting point to a target and continuously assigns new targets to the agents. Existing methods tend to encounter issues such as deadlocks and the need for replanning as the map size increases or the number of agents grows. This paper proposes a method combining guided channel constraints with congestion-aware communication. By applying the concept of guided channels and encouraging agents to move in the same direction, the complexity of the problem is reduced. The method also incorporates a congestion-aware communication mechanism, combining both global and selective communication to guide agents away from congested areas. This reduces communication frequency while improving the success rate of task completion. Experimental results demonstrate that, with this model, as the map size increases, the runtime is significantly shortened, and the throughput decay becomes less pronounced. Deadlocks and rerouting are greatly reduced, and the success rate is high.