International Workshop on Massively Multi-Agent Systems (2004.12. 10-11)

Challenges in Building Very Large Teams

Paul Scerri

Robotics Institute, Carnegie Mellon University

When agents coordinate according to the principles of teamwork they can flexibly, robustly and reliably achieve complex goals in complex, dynamic and even hostile environments. An emerging standard for building such teams is via the use of \emph{proxies}, which encapsulate domain independent coordination algorithms in a software module that works closely with a domain specific agent and other proxies to create a team. While succesful, previous generations of proxies and teamwork algorithms were limited to small teams because of their reliance on accurate models of the team and task state. By developing new algorithms that rely on probabilistic models we have been able to build teams several orders of magnitude larger than before. For example, a novel task allocation algorithm uses estimates of the availability of team members to performs a task, derived from probabilistic models of the team's capabilities, to allocate tasks extremely efficiently. These new scalable teamwork algorithms are encapsulated in the next generation of proxy and have been successfully demonstrated in several domains. However, key challenges remain before such teams can be deployed in real-world environments, including the need for languages to specify plans for such teams and ways of modeling and predicting team performance in new domains.