Research On Multi-agent Multilateral Negotiation Of Personalized Product Supply Chain Based On Q-learning

With the change of consumer demand,the production of personalized products is gradually becoming the core of enterprise development,and the improvement of Internet,information technology and manufacturing capabilities provide a material basis for the production of personalized products.The policy background of �Internet + Advanced Manufacturing� and �Supply Chain Innovation and Application� also provides policy support for the production of personalized products.Supply chain management under the background of Internet and information technology is a kind of platform management with the core of satisfying users' needs,which is more conducive to the development of personalized product supply chains.In order to promote the development of personalized product supply chain,this paper mainly studies from the following points:(1)Analyzed the current situation of supply chain development,analyzed the main reasons for the development of the personalized product supply chain in detail,and compared and analyzed the main characteristics of the personalized product supply chain,thereby constructed the operation mode of the personalized product supply chain,which is helpful to supplement and improve the relevant theory of personalized product supply chain.(2)Multi-Agent theory and reinforcement learning theory are important theoretical foundations for the study of supply chain negotiation.This study combined Multi-Agent theory and reinforcement learning theory to explore multilateral negotiation,breaking through the traditional bilateral negotiation model,and providing a convenient and efficient negotiation model for multilateral cooperation.(3)In the Multi-Agent multilateral negotiation model of personalized product supply chain based on Q-learning,this paper considered the impact of the negotiation participants' risk preference on negotiation progress,the loss of negotiating participants and negotiation utility,so as to make the negotiation model constructed in this study closer to the real negotiation scenario.(4)Construct fuzzy judgment conditions based on fuzzy theory.Compared with traditional negotiation judgment,fuzzy judgment can effectively improve the negotiation efficiency,and to a certain extent compensates for the shortcomings of the slow convergence of the Q-learning algorithm.Finally,a simulation experiment is carried out to verify the validity of the model.The results of the simulation experiment show that the risk preference of negotiating participants will affect the negotiation number,the concessions of negotiating participants and the negotiation utility.The higher the degree of risk preference of the negotiating participants,the better the convergence of negotiation model,the fewer times the negotiation agreement,the greater the negotiation loss,and the lower the negotiation utility.Conversely,the lower the degree of risk preference of the negotiating participants,the more consistent the number of times,the smaller the negotiation loss and the higher the negotiation utility.In addition,simulation experiments have found that in multi-objective negotiation,the higher the degree of risk preference of consumers in negotiation,the greater the concession in negotiation,the higher the price of negotiation and the lower the quality requirements negotiated.Finally,the study found that fuzzy judgment can effectively improve the efficiency of negotiation.This paper researched on Multi-Agent multilateral negotiation of personalized product supply chain based on Q-learning,which will help to achieve coordination and control between consumers,integrators and suppliers,making the whole consumption process more efficient,flexible and convenient.Maximize the benefits of the entire supply chain and improve the operational efficiency of the personalized product supply chain.