Task Decomposition and Role Sharing for Real-time Human-AI Swarm Collaboration

In spite of the impressive advances in artificial intelligence (AI), close collaboration between humans and AI systems is still difficult to achieve. To overcome this problem, we designed AI agents with a behavior tree that enables us to know what they are trying to do, and by using a consensus building algorithm, that is, a contract net protocol, a human and a group of AI agents were put together as one team. Taking advantage of this architecture, we designed an approach to decomposing cooperative tasks into appropriate roles. The effectiveness and feasibility of this approach were evaluated with teams in a simulated Tail Tag game. Matches were held with up to 29 AI agents and 1 person on one team and 30 people on the other team. The results indicate that our approach works almost evenly with human-human collaboration by sharing roles between a human and AI swarm. By understanding the roles of AI agents, a person can immediately understand the role that he/she should take. For further improvement, we also identified that it is necessary for a person to be able to give concise and global instructions.