Seminar 27th July 2010 1 p.m. University of Southampton Building 67 Room 1003
Information Management in Large Teams - Algorithms and Applications
Professor Paul Scerri
Carnegie Mellon: The Robotics Institute
- Web page
- http://www.ecs.soton.ac.uk/seminars/568
- Submitter
- Petrina Butler
Carnegie Mellon: The Robotics Institute
Abstract
Emerging, large heterogeneous teams working in complex environments will collect huge amounts of data. Any of that data might be useful to another member of the team, but most will not. Since it is infeasible to share all collected information with the rest of the team, agents must make local decisions about which information to share. In this talk, I will present the results from several analysis and algorithm design efforts that expose interesting emergent phenomena when information is shared in large teams. One of those emergent phenomena, that the reliability with which a team fuses uncertain information into the correct binary conclusion dramatically peaks when the frequency distribution of the size of information cascades is scale invariant, is exploited for a very efficient fusion algorithm. The analysis has also led to a new way of modeling multiagent systems that may lead to more efficient ways of designing such systems. Managing information is one of the keys to a number of applications we are developing. I will present several applications for which effective management of information is critical. Those applications include one that uses teams of unmanned aerial vehicles to locate radio signals and another that uses teams of robots to search for victims in an urban search and rescue environment.
Speaker Biography
Professor Paul Scerri is an Associate Research Professor at Carnegie Mellon University's Robotics Institute. His interests are in multi-agent and multi-robot systems, human robot interaction and adjustable autonomy. Professor Scerri has worked on and led several projects developing either cooperative agents or robots or human-robot interfaces. He has nearly 150 publications on these topics including in major conferences and journals. He is the primary designer and developer of the Machinetta software infrastructure for rapidly creating large, heterogeneous teams. Machinetta has been used in several projects for the Army, Air Force and Navy as well as for disaster response. Professor Scerri received his B.Sc. from RMIT University in Australia, his Ph.D. from Linkoping University in Sweden. Before moving to CMU, Professor Scerri was a post-doc at the University of Southern California.