Strategies for Querying Information Agents

Abstract

In a simple cooperative MAS model where a collection of "querying agents" can send queries to a collection of "information agents", we formalize the problem of designing strategies so that the expected completion time of the queries is minimized, when every querying agent uses the same strategy. We devise a provably optimal strategy for the static case with no query arrivals, and show via simulations that the same strategy performs well when queries arrive with a certain probability. We also consider issues such as whether or not the expected completion time can be reduced by sending multiple copies of queries, or by aborting copies of answered queries.