2018 Volume 31 Issue 4 Pages 167-176
When multi-agent systems explore in actual environments, it is difficult to manage agents centralizedly from outside due to communicable range. This paper proposes a maze exploration algorithm for distributed control system considering communicable range of agents. In proposed algorithm, agents share information each other within communicable range and then each agent detects deadlocks generating a wait-for graph from the shared information. Sharing cycle of wait-for graphs enables agents to take over the explorations each other and resolve deadlocks. This paper verifies the effectiveness of the proposed deadlock resolution algorithm via numerical experiments about some of communicable ranges and numbers of agents of exploration system. The simulation results show the influence of the proposed deadlock resolution algorithm on the efficiency of exploration and the path length of agents. The simulation results also show the relationship between the communicable range, the number of agents of exploration system and the failure probability of exploration.