Research On Strategy Of Multi-Agent Argumentation-based Negotiation Based On Self-learning

In the research on argumentation-based automatic negotiation based on agent,how to improve agents’self-learning ability during the process of argumentation-basednegotiation to exploit the characteristic of self-adaptability of agent is a urgent task.This paper makes a deep study of relevant problems about strategies withself-learning ability in argumentation-based negotiation. Firstly, the paper describesresearch situation and defects of self-adaptive strategy in argumentation-basednegotiation both at home and abroad, introduces basic theory involved in strategy withself-learning ability of argumentation-based negotiation.Secondly, on the basis of indicating relevant parameters and suppositions, thispaper discusses generation of self-adaptive concessional strategy, and two generatingmodels of concession strategy are also studied as: model based on time constraint andmodel based on opposite’s preferences with time constraint, the process ofdemonstration has been shown that the latter designed by PSO-RBFNN (RBF NeuralNetwork optimized by Particle Swarm Optimization) algorithm has better abilities oflearning and reasoning, which is dominant strategy in bilateral negotiations. Then, theretention value of issues involved in model based on opposite’s preferences with timeconstraint are analyzed and learned through the method of CBR(Case-basedReasoning), which will help negotiator agent to determine the concessional range ofissues, and further optimize core concessional strategy presented in this paper.Thirdly, on the basis of indicating main types of argumentaion, takingconcessional strategy (optimized) based on opposite’s preferences with time constraintas premise, this paper proposes interactive language and protocol suitable in researchon strategy of this thesis, presents a method of generating and selecting argumentationbased on the optimized concession strategy and judgment of conflict, and designs adialogue of argumentation to explain process of argumentation in detail, whichverifies rationality and validity of the method. Admittedly, strategy inargumentation-based negotiation is analyzed and studied through adopting appropriateprotocol, method and process, which has important academic significance andapplication value.Finally, this paper assesses rationality, stability and validity of strategy withself-learning ability in argumentation-based negotiation proposed in this research. A PSO-RBFNN is trained and tested to fit opposite’s preferences, and the network iscarried out stability evaluation through using algorithm of Bootstrap. Then, on thebasis of simulated data, retention value of issues are analyzed and learned through themethod of CBR to optimize core function of concessional strategy. Next, theefficiency of strategy is evaluated through comparative experiment. Afterwards, aframework of internal structure suitable in research on strategy of this thesis ispresented and a prototype system is finished preliminarily to realize the process ofbilateral argumentation-based negotiations based on multi-agent.This paper studies and disscusses related problems that involved in introducingmethod of self-learning into strategy of argumentation-based negotiation based onmulti-agent, proposes complete method of generation, optimization and selectionabout strategy with self-learning ability in argumentation-based negotiation, andcarries out theoretical exploration for constructing argumentation-based negotiationsystem with self-learning ability. Both theoretical analysis and test show that thestrategy proposed in this paper is of benefit to improve self-adaptive ability andintelligent degree of negotiator agent, which will promote business negotiation basedon multi-agent towards the development of high efficiency and win-win situation.