Multi-agent priority gaming for path planning in chain-like blocking situation

Prioritized planning is an important and efficient method for multi-agent path finding. During the solution process, the method may encounter the problem of high-priority agents blocking low-priority agents, and there remains a lack of an efficient method to address a series of chain-like blocking problems. This article proposes a multi-level conflict game method for mediating the priority conflicts of blocked agents. A level structure is designed to reflect the hierarchical relationship of chain-like blocking agents, and a multi-level conflict model is constructed through conflict escalation. Priority games on agents within levels cause high-priority agents to make way for low-priority agents. Priority game operations are carried out level by level, from high to low level, to achieve conflict de-escalation and mediation for all agents. Results of random tests on different scenarios demonstrate the effectiveness of our method in solving chain-like blocking problems, as well as its advantages of solving problems in dense scenarios with hundreds of agents within seconds and maintaining a success rate of over 88.0%.