Speaker: Professor Milind Tambe
Time: Wednesday, February 3, 6-8 pm
Location: USC’s Robert Zemeckis Center for Digital Arts (RZC)
Room 201 Zemeckis Media Lab (ZML)
Multiagent systems: Lessons learned from putting theory into practice
Milind Tambe
Teamcore research group
http://teamcore.usc.edu/tambe
http://teamcore.usc.edu
How do we build multiagent systems? Today, within the agents and multiagent systems community, we see four main approaches: logic-based belief-desire-intention (BDI), decision- theory and its incarnation in distributed markov decision problems (distributed MDPs or POMDPs), distributed constraint optimization (DCOPs) and finally, auctions or game-theoretic approaches. I will begin my presentation by providing some historical context of the field of agents and multiagent systems and an overview of these approaches.
In general, while there is exciting progress in this research, we still lack sufficient testing of our theories in complex multiagent domains, to evaluate their promised strengths and uncover unanticipated limitations. In this context, I will outline lessons learned in the Teamcore group’s recent efforts to transition theory into practice. I will focus in particular on research based on game-theory for randomizing plans for security applications and to avoid predictability that may be exploited by an opponent. Our algorithms are at the heart of ARMOR, a software scheduler that randomizes police checkpoints and canine patrols, deployed at the Los Angeles International Airport since August 2007. Our algorithms are also in use by the Federal Air Marshal Service for scheduling air marshals on limited sectors of international flights; and the Transportation Security Administration (TSA) is evaluating these algorithms for maintaining airport security at Pittsburgh and LAX airports towards possible large-scale deployments. I will also briefly discuss our research on DCOPs and their application to mobile sensor nets, and outline some recent research thrusts including multiagent-based evacuation simulation.
