Performance of heterogeneous robot teams with personality adjusted learning

This paper presents a reinforcement learning algorithm, which is inspired by human team dynamics, for autonomous robotic multi agent applications. Individual agents on the team have heterogeneous capabilities and responsibilities. The learning algorithm assigns strictly local credit assignments to individual agents promoting scalability of the team size. The Personality Adjusted Learner (PAL) algorithm is applied to heterogeneous teams of robots with reward adjustments modified from earlier work on homogeneous teams and an information-based action personality type assignment algorithm has been incorporated. The PAL algorithm was tested in a robot combat scenario against both static and learning opponent teams. The PAL team studied included distinct commander, driver, and gunner agents for each robot. The personality preferences for each agent were varied systematically to uncover team performance sensitivities to agent personality preference assignments. The results show a significant sensitivity for the commander agent. This agent selected the robot strategy, and it was noted that the better performing commander personalities were linked to team oriented actions, rather than more selfish strategies. The driver and gunner agent performance remained insensitive to personality assignment. The driver and gunner actions did not apply at the strategic level, indicating that personality preferences may be important for agents responsible for learning to cooperate intentionally with teammates.