Autonomy Negotiation Model Optimal Study In MAS

Nowadays, with the rapid development of e-commerce, it is coming intobeing a market with enormous potential. With the exploded increase of theinformation and the commerce, and the increasingly complexity of the networkenvironment, and the complex relationship of product provider and saler on theinternet, a kind of new technique of harmonizing the commerce relationbecomes more and more urgently to satisfy the demand of e-commerce. On theother hand, along with the developing requirement of globalation, individuationand betimes of electro commerce, the behavior of trade-off online is becomingmore and more complex, and the commerce treaty appears more frequently, andthe needs of autonomy negotiation is more and more high.Autonomy negotiation model is the base of binary Agent achievingmulti-issues autonomy negotiation. But practically, in the process of binarymulti-issues autonomy negotiation, the negotiation deadlock that is caused byone of the negotiation topics not achieved the balance point often occurs. Toimprove negotiation successing rate and negotiation use, the study of negotiation deadlock mechanism is necessary. And to promote the intelligent electroncommerce to practicality has important meaning. The resolution of negotiationdeadlock has becoming the important study task of optimizing the autonomynegotiation model.Aiming at the negotiation deaklock appeared in the negotiation process ofbinary Agent, and using the equivalent permutation of vectors, on the conditionof guaranteeing the unite avail not depress, use Q-learning method and Optionhierarchical reinforcement learning method to learn the vector of topicsreservation value dynamically. Clear up the deadlock to make the negotiationsuccess and improve the negotiation success rate and negotiation avail and reachthe purpose of optimizing the binary multi-issue autonomy negotiation model.Aiming at the analyse to start the task, the paper can be divided into sevenparts:(1) Expound the study background and the meaning of the paper, andsummarize the study actuality of the autonomy negotiation, the deadlockresolution in the negotiation and the hierarchical reinforcement learning.(2) Introduce the definitions and correlation concepts of Agent technic andMAS technic and the knowledge in multi-agents negotiation.(3)Expatiate the theory knowledge of Q-learning in the reinforcementlearning and hierarchical reinforcement learning. Summarize the automaticaldelamination problem in the hierarchical reinforcement learning which is themethod of establish the Option automatically. (4)Give out the binary multi-issue autonomy negotiation model and statethe proposal strategy and the negotiation protocol briefly.(5)Introduce both the Q-learning and the Option hierarchical reinforcementlearnig methods into the autonomy negotiation model described in the formerchapter. Clear up the deadlock and realize the optimization of the negotiationmodel.(6)Use the Q-learning method to sovle the deadlock appeared in the binarymulti-issue negotiation model, and programme with the C++ language, andsimulate with the Visual studio 6.0 to check up the validity of the method.(7)Summarize the work generally and put forward the shortage of the paper.Bring forward the possible study direction.