不对称信息下项目调度和资源规划的多方协同优化研究

Multi-party collaboration is inevitable for specialized social division and delicacy management, and the effective decision and common interest are management difficulty and research focus. The research on decentralized multi-projects and project-driven supply chain did not consider information asymmetry, uncertainty and contract flexibility. Therefore, only the resource information required by the projects is open. The lying and bargaining for self-interest exist during the competition, and there are also the integrity mechanism and incentive contract for collaboration. Moreover, besides uncertain durations, project decisions cannot estimate accurately others’ uncertain time and relevant decisions. Besides, in the collaborative optimization of project-driven supply chain, incentive strength changes with the ordering conditions and game severity. We will present multi-agent collaboration mechanism, and form a closed loop with multi-objective competitive evolutionary algorithm, in order to promote algorithm performance and make the satisfactory solution feasible under information asymmetry. The research will connect the project chain and resource chain seamlessly and respond to the market risks agile. The research will finally build a cooperation league, and provide decision supports for collaborative decision and intelligent management.

分工专业化、施工精细化必然会导致多参与方协同合作，而在信息不对称下如何促进多方利益一致性和决策有效性是管理难点，也是研究热点。本研究根据资源属性不同，对分散式多项目调度和项目驱动型供应链分别展开研究。目前研究中一般不考虑信息不对称性和不确定性。因此，本研究将在协同优化过程中仅公开资源需求信息，允许参与方在竞争中撒谎、讨价还价发展自身利益，总控企业也将通过诚信机制、激励契约促进协作。除活动工时不确定外，决策者还不能精确估计其他参与方的不确定性时间以及相关决策。另外，契约激励强度可随订购环境、双方博弈程度而变化。本研究将提出相应的多Agent协同机制，并将其嵌入竞赛式多目标进化算法中形成闭环，既提高算法性能，也保证满意解在信息不对称下可行。本研究将分析参与方博弈策略影响关系，促进企业项目链和资源链无缝连接，在市场竞争中敏捷应对风险，最终建立多方协作联盟，为项目协同决策、智能管理提供理论支持。

经济全球化发展中专业分工有利于提高社会生产力，而各主体在信息不对称下如何有效的管理项目链、供应链以达到合作共赢是研究难点。对于带有全局资源的项目链，本研究在考虑资源异质性和信息不对称下，提出组合多智能体并设计算法来解决分布式多项目调度问题，探究了项目经理的谎报策略与规划者的诚信政策对系统性能的影响。研究发现在多数情况下，规划者的诚信政策可减少项目智能体的谎报行为对全局目标的负面影响；同时通过与中心协调方法的对比，发现组合多智能体系统在搜索性能和计算时间上均优于中心协调方法。对于项目驱动供应链，本研究考虑不同契约对供应商提供积极或消极的激励，分析了四种类型激励契约对项目经理以及供应商决策的影响；提出了多智能体和进化算法相结合的混合算法。研究发现在正常风险情况下，采用非财务激励契约对项目经理和供应商都有利，而在工期紧张的情况下，采用惩罚对双方更有利，该结论可以指导项目经理选择合适的激励合同来保障项目的实施进度；同时通过对混合算法和遗传算法对比实验，发现两种算法之间的差距非常小，并且随着项目规模的增长，差距明显减少。