This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Predictability & Criticality Metrics for Coordination in Complex Environments Rajiv T. Maheswaran Information Sciences Institute Univ. of Southern California 4676 Admiralty Way, #1001 Marina Del Rey, CA 90292 [email protected] Pedro Szekely Information Sciences Institute Univ. of Southern California 4676 Admiralty Way, #1001 Marina Del Rey, CA 90292 [email protected] M. Becker, S. Fitzpatrick, G. Gati, J. Jin, R. Neches, N. Noori, C. Rogers, R. Sanchez, K. Smyth, C. Vanbuskirk ABSTRACT We address the problem of coordinating the activities of a team of agents in a dynamic, uncertain, nonlinear environment. Bounded rationality, bounded communication, subjectivity and distribution make it extremely challenging to find effective strategies. In these domains it is difficult to accurately predict whether potential policy modifications will lead to an increase in the value of the team re- ward. Our Predictability and Criticality Metrics (PCM) approach errs on the side of safety, and advocates considering policy modifi- cations that are guaranteed to not harm the current policy, and uses simple metrics to choose from within that set a modification that increases the team reward. In the context of the DARPA Coordi- nators program, we show how the PCM approach yielded a system that significantly outperformed several competing approaches in an extensive independent evaluation. General Terms Algorithms Keywords Multi-Agent, Uncertainty, Dynamism, Coordination, Scheduling 1. INTRODUCTION The coordinated execution of activities of a multi-agent team in dynamic and uncertain environments is of critical interest in do- mains such as large-scale disaster rescue, joint military operations and project management, among others. There are many character- istics of these domains that make effective coordination extremely challenging. The team begins with an initial plan of activities which have uncertainty in duration and outcome. As uncertainties are re- solved through execution, agents may need to modify their plans, e.g., change timings or perform alternate activities. As the scale of the problems increases, it becomes infeasible to calculate and store an optimal set of policies that prescribe appropriate plan changes for all contingencies. This introduces one form of dynamism , where agents must modify their policies over time. A second form of dy- namism occurs when agents’ models of the world or the team re- ward function changes during execution. An agent might discover Cite as: Predictability & Criticality Metrics for Coordination in Complex Domains, R. T. Maheswaran, P. Szekely, M. Becker, S. Fitzpatrick, G. Gati, J. Jin, R. Neches, N. Noori, C. M. Rogers, R. Sanchez, K. Smyth, and C. VanBuskirk , Proc. of 7th Int. Conf. on Autonomous Agents and Multiagent Systems (AAMAS 2008) , Padgham, Parkes, Müller and Parsons(eds.),May,12-16.,2008,Estoril,Portugal,pp.647-654....
View Full Document
This note was uploaded on 04/19/2010 for the course CISC 800 taught by Professor Kd during the Spring '10 term at University of Delaware.
- Spring '10