RECON: A robust multi-agent environment for simulating concurrent negotiations

Abstract

Recon is an experimental simulation platform that supports the development of software agents interacting concurrently with other agents in negotiation domains. Unlike existing simulation toolkits that support only imperative negotiation strategies, Recon also supports declarative strategies, for applications where logic-based agents need to explain their negotiation decisions to a user. Recon is built on top of the GOLEM agent platform, specialized with a set of infrastructure agents that can manage an electronic market and extract statistics from the negotiations that take place. We evaluate the performance of Recon by showing how by increasing the number of agents in a simulation affects the agents’ time to make an offer during negotiation.