Learning Models of the Negotiation Partner in Spatio-temporal Collaboration

Abstract. We describe an approach for learning the model of the opponent in spatio-temporal negotiation. We use the Children in the Rectangular Forest canonical problem as an example. The opponent model is represented by the physical characteristics of the agents: the current location and the destination. We assume that the agents do not disclose any of their information voluntarily; the learning needs to rely on the study of the offers exchanged during normal negotiation. Our approach is Bayesian learning, with the main contribution being four techniques through which the posterior probabilities are determined. The calculations rely on (a) feasibility of offers, (b) rationality of offers, (c) the assumption of decreasing utility, and (d) the assumption of accepting offer which is better than the next counter-offer.

• Type: Conference
• Year: 2008
• Where: COLCOM
• Authors: Yi Luo, Ladislau Bölöni
• Keywords: Applied Computing, COLCOM 2008, Forest Canonical Problem, Opponent Model, Spatio-temporal Negotiation