节点文献

协作智能体环境下虚拟企业自由联盟构建技术研究

Study on Constructing Virtual Enterprise Free Alliance under Multi-agent Collaborative System

【作者】 宁红云

【导师】 刘金兰;

【作者基本信息】 天津大学 , 信息管理与信息系统, 2009, 博士

【摘要】 虚拟企业是以计算机网络为支撑的跨地理空间进行信息传递和管理的临时企业组织形式,它已成为二十一世纪企业追求的重要运作模式之一。本文在广泛查阅国内外文献的基础上,从分析网络化企业联盟所面临的困难着手,深入研究了面向实时协作的网络化虚拟企业运作的系统理论框架,给出了网络化虚拟企业运作模型,研究涉及企业交互协作建模,协同数据管理与事务处理,伙伴在线选择与定位协商等关键问题。论文取得的主要成果和创新点如下:⑴基于协作智能体在分布计算环境中的可变结构特性,提出网络化虚拟企业运作的系统理论框架,引入动态域的管理模式、动态业务流协作模型以及多智能体角色控制机制,有效解决了动态联盟中多边业务的动态规划与协调问题,从而提高了虚拟企业面向任务进行重组的灵活性,增强了系统的适应能力。(2)针对虚拟企业协作数据访问特点,引入改进的协作对象及其虚拟表机制及智能体移动策略,提出基于三层结构、四级模式的分布式协作数据管理模型,引入视图自维护思想分别从保存有限实视图,为特定数据开辟缓存和实时监控同根数据等方面,有效改善了虚拟企业中动态业务流数据的协作访问性能。(3)针对传统方法在虚拟企业动态数据管理中事务错失率高和难以适应动态事务处理的缺陷,结合线程与协作智能体并行处理技术,提出基于动态域的事务3阶段处理法,以及动态多版本乐观并发控制和协作查询优化策略,有效降低了动态域事务的错失率和重启率,并提高了事务处理能力。⑷从分析网络化虚拟企业运作模式在伙伴选择中所具有的优势着手,提出一种基于实时协作关联图的虚拟伙伴在线自动选择模型和基于蚁群优化的潜在伙伴挖掘模型,新模型与传统的离线挖掘方法相比具有更好的实时性和准确性,同时有效解决了企业合作中的间接关联和弱关联的挖掘问题。(5)针对传统虚拟企业沟通耗时和协商失败率高等问题,为满足分布式网络环境下快速构建企业联盟的目标要求,基于同步约束的多边伙伴定位协商协议,提出伙伴定位在线协商模型及其协进化协商决策算法,它通过实时更新智能体信念明显地改进了协商行为的效用,有效提高了伙伴定位在线协商的成功率。

【Abstract】 Virtual Enterprise (VE) is a cross-geospatial provisional organization supported by information delivery and management on internet. It has become the pursuit of the important operation mode for enterprises in the 21st century. By reading the literature domestic and abroad, we carefully analyzed the problems of network-based enterprise alliance, deeply studied the theory architecture about real-time collaborative VE network system, and then proposed a novel networked VE operation model, which involves some key problems of cross-enterprise collaboration modeling, collaborative data management and transaction processing, partner online selection and negotiation, etc. The main content or possible creativities of the dissertation are as follows:(1) Based on variable structure characteristic of multiple collaborative agents under distributed computing environment, the theory architecture about VE network operation system is presented. With the introduction of dynamic domain-based organization-mode, workflow collaboration model and role-control mechanism, the system effectively solves the dynamic programming and coordination problems of multilateral operation to improve the VE task-reorganization flexibility. These features also enhance the adaptability of the system.(2) In view of VE cooperation data accessing characteristic, a novel distributed cooperation data management model with 3-tier structures and 4-level modes is proposed, which adopts the collaboration object and its virtual table mechanisms, as well as the mobile-agent strategy. Base on the idea of view self-maintained, technologies of saving limited materialized view, caching specific data and real-time monitoring data of same root, are provided. The model increases the VE cooperation data-access performance to meet the dynamic domain-based workflow processing.(3) To solve the problem of high transaction miss-rate in traditional VE data management, which is difficult to adapt to the dynamic situation, a dynamic domain-based 3PP-Method and some strategies for dynamic multi-version optimistic concurrency control and cooperation inquiry optimization are proposed, combining the threading technology and the collaborative agent parallel processing technology. 3PP-Method effectively reduces the domain-transaction miss rate and restart rate, as well as improves the processing capacity of the domain-transaction.(4) Taking the advantage of partner selection on VE network operation system, a new dynamic mining model for partner online selection and an ACO-based potential partner mining model are provided, based on a real-time collaboration incidence graph. Comparing to the traditional offline mining methods, these models have better real-time accuracy and the indirect or weak associations in business cooperation also can be effectively caught.(5) For the problem of high failure-rate and time-consuming process of traditional consultation, aiming to construct VE rapidly under distributed network environment, a novel virtual partner-location online-negotiation model with the co-evolution algorithm is proposed based on the multilateral negotiation agreement of synchronized restraint, which improves the utility of negotiation behavior by helping agents duly update their belief in real-time. Thus, the success rate of virtual partner-location online-negotiation is obviously promoted.

  • 【网络出版投稿人】 天津大学
  • 【网络出版年期】2010年 12期
节点文献中: