Research On Bilateral Multi-issue Negotiation Model For A Kind Of Complex Environment

Bilateral multi-issue negotiation is an important research area of automated negotiation.Negotiation in complex environments has been the hotspot in automated negotiation.Since the introduction of large-scale outcome space and time constraint in multi-issue negotiation,so that it is one of the challenging problems to establish an efficient negotiation model in complex environment.This paper presents a negotiation model based on the Gaussian process regression to predict the opponent's concession,which change risk attitudes through dynamic risk functions.This paper presents two types of dynamic risk strategies.The first strategy is based on negotiation time.Agents change the risk attitude according to the elapsed time in the negotiation process,and then calculate the risk of opponent by analyzing the opponent's concession strategies.Agents will not accept opponent's with low utility by this way.The second strategy is based on the utility threshold.Agent's risk atitude depends on the diffirent between the utiltiy of opponent's offer and our utility threshold.Based on the prediction of opponent's concessions,our Agents change the risk attitude according to the utility of opponent's offer.According to the dynamic risk strategy,we implement the Agents in GENIUS.The negotiation model based on the above strategy implements in the tournament GENIUS negotiation platformand negotiates with other Agents.Empirical results show that the Agent in this paper can adapt to different negotiation domain under the complex environment.The dynamic risk strategy based on negotiation time can adjust the risk factors to change the Agent's attitudes,which according to consuming time of negotiation and value of opponent's offer.However the negotiation strategy based on utility threshold,through the utility of opponent's offer can decide its own risk strategy and then change the concession rate.Two strategies for bilateral multi-issue negotiation in the different domain can obtain higher utility,and consume less time,that verified the negotiation model which using dynamic risk mechanism in the paper can make good performance in various types of negotiations domains.