基于多Agent制造业供应链产销协同冲突协商研究

The implementation of production-marketing-collaboration in manufacture has been a trend. It can be effective to deal with complex market competition, rapidly meet customers' dynamic demand, reduce the operational risk and the production cost, increase the profit of enterprise. However, conflicts often appear during the production-marketing-collaboration process. If the conflicts can not be promptly eliminated, the cooperation and development of the enterprises would be obstructed. Negotiation is an effective way to solve this kind of conflict. But the traditional negotiation is more and more difficult to adapt to the new environment - the digital world, and it always leads to the poor implementing effect. Negotiation based on multi-agent is a new method that can replace human to negotiation automatically, cover the shortages of the traditional negotiation, and improve the negotiation effect. However, at present, the researches of conflict negotiation based on multi-agent lack system integration, and disconnect with the practice. This project intends to take advantage of distribution, interaction and intelligence of multi-agent system, fuse manufacturing supply chain information resources and support the negotiation decision-making. We will design a negotiation system framework based on multi-agent system, research the negotiation decision-making, modeling, optimization strategy and adaptive concession strategy based on argumentation, so as to improve the rationality, timeliness and effectiveness of the negotiation and put forward a new train of thought for solving the conflict problems of the production-marketing-collaboration in supply chain. This study is beneficial to improve the intelligence level of negotiation and promote the coordinated development of the supply chain。

制造业供应链实施产销协同是一种发展趋势。它可以有效应对复杂市场竞争，快速满足客户动态需求，降低运行风险和生产成本，增加企业利润。然而在产销协同过程中，时常会产生冲突。消极的冲突如果得不到及时消解，会阻碍协同发展。协商是解决冲突有效途径。而传统协商，越来越不能适应数字世界新环境，往往导致协商效果差。 基于多Agent的协商是一种可以代替人进行自动协商的新方法，能够弥补传统协商不足，改善协商效果。然而，目前基于多Agent冲突协商研究缺乏系统整合，与实践脱节。本项目拟利用多Agent系统的分布、交互和智能性优势，融合制造业供应链协同信息系统资源，支持协商决策；设计一种基于多Agent协商系统框架，研究其协商过程、建模方法、决策优化以及基于辩论的自适应让步策略。从而提高协商的合理性、及时性和有效性，为解决供应链产销协同冲突问题提出一种新思路。本研究有利于提高协商智能化水平，促进供应链协同发展。

制造业供应链实施产销协同是一种发展趋势。它可以有效应对复杂市场竞争， 快速满足客户动态需求，降低运行风险和生产成本，增加企业利润。然而在产销协同过程中时常会产生冲突。消极的冲突如果得不到及时消解，会阻碍协同发展。协商是解决冲突有效途径。而传统协商，无法适应数字世界新环境，往往导致协商效果差。. 基于多 Agent 的协商可以代替人进行自动协商，能够弥补传统协商不足改善协商效果。本项目在产销协同过程中，针对上下游企业间存在信息差异，而产生冲突的问题，对完全信息共享和有限信息共享两种情形，基于Agent的纳什协商，研究了协商冲突点的确定方法，并提出了一种基于Agent的启发式协商方法进行冲突消解。同时关注上下游市场需求不断变化，制定适应对手提议及市场需求变化的智能协商策略，建了基于黑板机制的多Agent产销协同协商框架及协商流程，设计了三种基于优化机器学习算法的自适应协商方法，给出了自适应协商策略。最后引入辩论因素进行协商，提高协商的灵活性和适应性，借鉴前景理论，给出了基于辩论的协商模型的元组表示、基于多Agent的辩论协商模型交互规则、基于前景理论的多Agent辩论协商机制，提出了基于前景理论的多Agent辩论协商模型和策略。还分别考虑了证据推理、供需双方的公平关切行为、关联议题发现以及考虑协商环境动态变化的协同优化策略。. 项目共发表CSSCI/CSCD/EI/SCI论文共24篇，其中：国家基金委认可管理类A刊4篇，教育部认可的管理类A刊3篇，基金委B类期刊1篇，SCI检索期刊1篇，EI检索5篇；完成专著1部，申请软著1项；团队3名青年教师均晋升副高级专业技术职务，1名青年教师到国外进行短期学习交流；16名研究生顺利毕业。. 项目针对制造业供应链产销协同中产生的冲突，突破传统的协商模式，结合优良的协商经验和知识，利用多Agent 系统，给出一种智能化的冲突协商方法和自适性协商策略。本研究成果有效促进协商决策的智能性、科学性，达到快速协商决策或自动协商的目的，大大提高制造业供应链协同协商智能化管理水平。