共享经济模式下考虑供需双方行为特征的分布式制造系统决策机理研究

This project mainly focuses on the distributed manufacturing system (DMS) under sharing economy scheme. In order to tackle the manufacturing unit distribution, resources allocation, markect coordination, and order matching in DMS operational management, based on the bounded rationality property of suppliers and customers in our model setting, tailored decision making supports are provided to each stage of DMS operating process. In the many server heavy load queueing network model, the unique property of DMS such as return of product and quality control unit in the system, together with capacitated lot-sizing is integrated, and the optimal manufacturing unit quantity is obtained w.r.t. time based on fluid model. When dealing with resource allocation problem in DMS, the system is decomposed into two stages, approximation algorithms are designed to obtain the order allocation policy. Besides, in the decentralized mode constructed by multi-agent framework, various non-cooperative game models are discussed. In particular, the fuzzy class of decision makers is established in the multi-stage game model with incomplete information, as an extension of the Bayesian game theory. Further, in consideration of the vaious types of evaluation formats given by decision makers, a inference rules and a learning mechanism based information aggregation methods are proposed within the environment of complex uncertain information. Finally, a comprehensive decision support system is developed to facilitate the decision making in DMS.

本项目面向共享经济模式下的分布式制造系统(DMS)，针对节点布局、资源调配、市场调节和订单处理等DMS运作过程中的关键决策类问题，基于供需双方均为有限理性决策主体的假设，分别在各阶段为其提供最佳决策支持。为提高DMS的服务水平以及核心制造资源的利用率，在多服务台重负荷排队网络模型中，集成DMS特有的返修退货与可变批量特征，基于流体模型求得DMS中制造单元的实时数量。在研究DMS资源调配时将系统分解为两个阶段，通过构造近似算法寻求集中式系统管理方式下的订单分配方法。此外，在DMS的多代理去中心化管理模式中，依据订单与DMS产能的关联，在代理间引入多种竞争型博弈模型，特别在包含动态不完全信息的博弈环境中建立决策主体的模糊类别，拟拓展经典贝叶斯博弈理论。考虑到决策主体评判形式的多样性特点，提出复杂不确定性信息条件下基于逻辑推理和学习机制的信息融合方法，并建立相应的的决策支持系统。

本项目面向共享经济模式下的分布式制造系统(DMS)，针对节点布局、资源调配、市场调节和订单处理等DMS运作过程中的关键决策类问题，基于供需双方均为有限理性决策主体的假设，分别在各阶段为其提供最佳决策支持。为提高DMS的服务水平以及核心制造资源的利用率 ，在多服务台重负荷排队网络模型中，集成DMS特有的返修退货与可变批量特征，基于流体模型求得DMS中制造单元的实时数量。在研究DMS资源调配时将系统分解为两个阶段，通过构造近似算法寻求集中式系统管理方式下的订单分配方法。考虑到决策主体评判形式的多样性特点，提出复杂不确定性信息条件下基于逻辑推理和学习机制的信息融合方法 ，并建立相应的决策支持系统。本项目在执行期内发表国内外高水平期刊论文三十余篇，发表学术专著一部，成果获得2020年中国商业联合会科技进步特等奖，2019年吴文俊人工智能科技进步二等奖，同时获得中国工程院2035咨询项目一项，国家自然科学基金委新立项面上项目一项，取得了显著的经济效益和社会效益。