Study, Based On The Similarity Of Mechanisms Of Negotiation Strategies

In recent years, bilateral negotiation strategy mechanism is an important area in the practical research on MAS (Multi-Agent System). In the research on negotiation strategy, dynamics and uncertainties are important issues. In the literature of negotiation strategy, there are three main research methods, known as game theory, heuristic approach and argumentation based negotiation.In the research of heuristic approach, Jennings et al. pointed out it would bring a lot of difficult into the design of negotiation strategy mechanism to introduce issue dynamics into negotiation because it would result in the changes of contract space. To solve this problem, Faratin et al. introduced issue manipulation mechanism into the proposal generation strategy mechanism and put forth a trade-off mechanism based on fuzzy similarity theory with the support of META strategy. In their research, issue dynamics are allowed, but there are some problems in their tradeoff algorithm.The negotiation process of MAS is detailed, the concept of negotiation is depicted and types of negotiation are detailed in this paper. Also, the dynamics and uncertainties are detailed separately.This paper focuses on how to guarantee the joint gain and the negotiation time under a bilateral negotiation background which incorporates the following properties: incomplete information, limited negotiation time and issue set can change dynamically. To achieve it, a generic bilateral multi-issue negotiation model is proposed at first, then a bilateral multi-issue negotiation utility evaluation mechanism based on multi-attribute utility theory is put forth and the trade-off algorithm based on fuzzy similarity is extended according to the relations between issues.In addition, a GA (Genetic Algorithm) based proposal generation strategy is analyzed, in which the limits on the weighted Euclidean distances of the fitness function are pointed out. To remove these limits, an improved GA based proposal generation strategy GAS (Genetic Algorithm Strategy) is proposed. The experiments show that the improved genetic algorithm GAI (Genetic Algorithm Improved) can give a better performance over non-line utility functions. The stability of the algorithms is analyzed by experiments.