An efficient automated negotiation strategy for complex environments

A complex and challenging bilateral negotiation environment for rational autonomous agents is where agents negotiate multi-issue contracts in unknown application domains with unknown opponents under real-time constraints. In this paper we present a negotiation strategy called EMAR for this kind of environment that relies on a combination of Empirical Mode Decomposition ( EM ̲ D ) and Autoregressive Moving Average ( AR ̲ MA ). EMAR enables a negotiating agent to acquire an opponent model and to use this model for adjusting its target utility in real-time on the basis of an adaptive concession-making mechanism. Experimental results show that EMAR outperforms best performing agents from the recent Automated Negotiating Agents Competitions (ANAC) in a wide range of application domains. Moreover, an analysis based on empirical game theory is provided that shows the robustness of EMAR in different negotiation contexts.