基于MAS的分布式突发事件应急风险决策方法研究

The risk characteristics of emergency decision is the critical factor to influence the decision outcome. According to the complex decision environment with multiple decision makers, multiple interrelated decisions and decision conflicts, the multi-agent system based distributed risk decision-making method for emergency response will be investigated..Firstly, the risk analysis and collaborative decision method of multi-agent’s emergency decision which supports the distributed risk decision network modeling and reasoning, the distributed decision objective management, commitment and generation, as well as the distributed scheme optimization, voting and selection will be addressed..Secondly, the conflict detection and resolution method of multi-agent’s risk based emergency decision which supports the distributed conflict detection of risk based emergency decision network, decision objective and scheme, as well as the parallel and multi-linked conflict resolutions will be proposed..Thirdly, the performance evaluation and dynamic optimization method of multi-agent’s risk based emergency decision which supports the information complexity of risk analysis, objective generation, scheme selection and conflict resolution, the collaborative reliability of decision maker, decision content, and decision process, as well as the dynamic optimization of information interaction and collaborative relationship will be presented..These research works will make effort to extend the theories and methods of risk based emergency decision and MAS based distributed decision, as well as to provide the scientific basis and theoretical guidance for the distributed and collaborative emergency decision practice.

突发事件应急决策的风险特性是影响决策效果的关键因素。针对动态不确定环境下的多决策者参与、多决策关联和决策之间冲突的复杂决策环境，围绕基于MAS（Multi-Agent System）的分布式突发事件应急风险决策方法展开研究，主要包括：①支持分布式风险决策网络建模与推理，决策目标管理、承诺与生成，备选方案最优化、投票与优选的多Agent应急决策风险分析与协作决策方法；②支持分布式风险决策网络、决策目标与决策方案冲突识别，并行与链式冲突消解的多Agent应急风险决策冲突检测与协调方法；③支持风险评估、目标形成、方案优选与冲突协调信息复杂性，决策主体、决策内容与决策过程协同可靠性，信息交互与协同关系动态优化的多Agent应急风险决策绩效评价与动态优化方法。力图实现应急风险决策与MAS分布式决策理论与方法的有力拓展，同时为分布协作式的突发事件应急决策实践提供科学依据和理论指导。

围绕考虑动态不确定环境下的多决策者参与、多决策关联和决策之间冲突的分布式突发事件应急风险决策问题，运用基于MAS（Multi-Agent System）的分布式决策问题建模，以及多Agent协作与自动协商理论，针对多Agent应急决策风险分析与协作决策方法，多Agent应急风险决策冲突检测与协调方法，以及多Agent应急风险决策绩效评价与动态优化方法展开研究。主要研究内容包括：.1）针对多Agent应急决策风险分析与协作决策方法目标，主要研究了基于BDN和Multi-Agent的突发事件应急风险决策方法、基于前景理论的风险型多属性应急群决策方法、考虑设施选址和资源/人员协同优化的灾后应急决策、考虑灾民分级的应急避难场所选址与资源分配决策、考虑救护车繁忙率的应急医疗设施选址与分配决策。.2）针对多Agent应急风险决策冲突检测与协调方法研究目标，主要研究了基于多Agent协商的多对多主体决策冲突协调方法、考虑依赖关系和能力约束的多对多决策主体任务分配方法、基于竞标和辩论的多对多主体决策冲突协调方法、能力约束情况下的供需双方任务分配博弈与协调分析。.3）针对多Agent应急风险决策绩效评价与动态优化方法研究目标，主要研究了基于HLA-MAS的分布式多主体决策建模与仿真方法、基于多主体和前景理论的应急风险决策仿真、危化品运输监测与预警决策仿真评估、危化品运输安全应急救援决策仿真评估、考虑领导者和恐慌情绪的多Agent人员疏散过程仿真方法。