Graphics enhanced version of this site
Paul Scerri
Systems Scientist
Email address: pscerri@cs.cmu.edu
Office: NSH 1617
Phone: (412) 268-2145
Mailing address:
Carnegie Mellon University
Robotics Institute
5000 Forbes Avenue
Pittsburgh, PA 15213
For more information, see my personal homepage.
Jump to: Research interests | Keywords | Projects | Publications
My research centers around the idea of coordination techniques based on the concept of /teamwork/. Teamwork requires benevolent team members act in particular ways towards each other, e.g., informing team mates of important events in the environment and committing jointly to team plans. In theory, teamwork promises the types of properties we desire in heterogeneous organizations, including flexibility, efficiency and robustness. However, while people may have well developed abilities to engage in teamwork, agents and robots typically do not. A key research issue is thus to give agents and robots the capability to perform as a part of a team. However, due to limitations in agent and robot sensing and reasoning it is likely that teamwork as a group of people might perform it may not be feasible for heterogeneous teams. Moreover, teamwork, as implemented by a human basketball or research team, may not be the most efficient way of coordinating highly heterogeneous entities. A long term aim of my research is to determine how teamwork should be implemented in a heterogeneous team.
Specifically, my research so far has focused on two aspects of fulfilling the vision of effective heterogeneous teams. The first thrust of my research has been on developing a reusable infrastructure for teamwork. The key idea is to provide each robot, agent and person with a /proxy/. The proxies manage the details of the teamwork, e.g., maintaining coherence, while interacting only at a high level with the robot, agent or person. We have made important advances in coordinating intelligent distributed entities within the framework of the proxies (which are publicly available). Use of the proxy-based infrastructure in two domains, a research group and urban disaster recovery, has show the potential of the idea, but also revealed challenging new research problems. For example, one issue revealed by the urban rescue domain is the need for new techniques for specifying and monitoring team plans, when highly heterogeneous entities will be executing the plans.
The second specific focus of my research has been on adjustable autonomy. With adjustable autonomy entities dynamically change their level of autonomy to best coordinate with others. The most extensive use to date of adjustable autonomy has been in the interactions between a proxy and its human user. In particular, the proxy must decide whether to make coordination decisions autonomously on behalf of the person or relinquish control and request input. The key to adjustable autonomy is balancing the desire to get high quality input (e.g., people know their own preferences best) while not overburdening users or causing miscoordination with the rest of the team. While we were not the first to look at adjustable autonomy, our work was the first to look at adjustable autonomy in a coordination context. The coordination context exposed weaknesses with previous work that we addressed with new techniques and formal models. Our formal models and decision theoretic techniques provide tools for researchers for a range of cutting edge applications. Much exciting work remains to be done in this area.
artificial intelligence, constraint-directed reasoning, human-computer interaction, and multi-agent systems