The Effect of Coalition Imbalance on Multi-Robot Teams

The notion of coalition imbalance refers to the variance in the individual contributions of team members towards the completion of a multi-agent (robot) task. Until now, the literature has studied imbalance purely from a fault tolerance perspective where more balanced coalitions were shown to be more robust to individual robot failure. However, the effect that balancing has on the overall performance of a multi-robot team still remains unexplored. This paper attempts to identify the relationship between the imbalance level of a multi-robot team and team performance. Experiments were conducted with a simulated multi-robot soccer team to demonstrate that teams lying at the extremities of the performance spectrum tend to have a higher balance level. Latter sections of the paper describe experiments that were conducted to demonstrate how balance information may be utilized to improve overall team performance.