Behavioral diversity in learning robot teams

Abstract

This work investigates the origins of behavioral diversity in learning robot
\nteams. Behavioral diversity refers to the extent to which agents assume
\ndistinct behavioral roles in a group. Most research in multi-robot teams to
\ndate has centered on homogeneous systems, with work in heterogeneous groups
\nfocused primarily on mechanical and sensor differences between agents. In 
\ncontrast, this work examines teams of mechanically identical robots. These
\nsystems are interesting because they may be homogeneous or heterogeneous
\ndepending only on behavior. Behavior is an extremely flexible dimension of
\nheterogeneity in learning teams because the agents converge to hetero- or
\nhomogeneous solutions on their own. This research provides new tools for the
\ninvestigation of behavioral diversity in multi-robot systems and a
\nsignificant body of results using these tools in simulated and real mobile
\nrobot experiments. The experiments specifically investigate the relationship
\nbetween the reinforcement function used for training and the diversity and 
\nperformance of the resulting multi-robot teams.