突发事件下应急车辆调度中的不确定风险评估与动态决策

Emergency vehicle dispatching problem is a complex decision-making problem with open multi-agent and multi-factor characteristics. The problem can be decomposed into four parts: risk assessment of the emergency, emergency vehicle scheduling, risk prediction and warning and dynamic dispatching. First, reasonable contingency requirements will be inferred from the risk assessment which are carried out by the emergency. Since the incident risk factors have high degrees of mutation, nonlinear, and correlated features, a more general measure satisfying subadditivity and duality properties will be used to carry out risk assessment. Based on the obtained emergency needs, we will build optimal vehicle routing models containing uncertain variables or uncertainty sets such as the optimal chance constrained model and multi-objective model. Furthermore, we will forecast the evolution risk based on uncertainty inference theory. A decision making process including "uncertain network"-"optimal vehicle dispatching"-"decision making application" will be proposed. Finally, hybrid intelligent algorithm will be designed to solve the mathematical models. The project aims to build multidisciplinary strategic management approach and to establish decision-making management methods with lacking of historical data and involving in experts' empirical data.

突发事件中的应急车辆调度问题是国际热门问题,具有多主体、多因素、多变性等特征。该问题的决策过程分解为突发事件的风险评估、应急车辆调度、风险预测预警、动态应急车辆调度等关键环节。本项目首先对突发事件进行不确定风险评估和预测，鉴于突发事件的风险因素有突变性、非线性和耦合性等特点，一种具有自对偶、次可加的不确定测度将被用来对风险进行评估，建立该测度下的应急风险指数概念，并研究其数学性质。在获取合理应急需求的基础上，研究含有不确定变量或者不确定集的最优车辆路径优化问题，建立最优应急车辆调度的机会约束模型和多目标不确定规划模型。基于不确定推理理论，研究应急车辆调度救援与突发事件演化的预测预警，建立基于"不确定网络-模型计算-车辆调度应用系统"的情景-应对决策方法。最后通过设计混合智能算法对模型求解。本项目拟构建包含多学科的战略管理方法，特别是在历史数据缺失情况下，构建专家经验参与的应急决策管理方法。

在突发事件应急决策中存在着大量的不确定性，含有不确定性的风险管理决策问题目前已成为国际该领域的研究热点之一. 本项目的创新之处在于：首先，针对应急决策问题中的不确定性本身固有的特征，利用自对偶、次可加测度刻画这些不确定性，研究刻画不确定性序列的极限性质和动态变化特点，分别建立基于自对偶、次可加测度的不确定多目标规划和目标规划的应急车辆调度模型；其次，通过模型分析，利用逆不确定分布和α-轨道性质建立了确定性等价模型，并将问题转化为多阶段离散动态规划模型和连续时间最优控制模型；结合“情景-应对”方案将不确定推理应用到多阶段决策中，结合决策者风险态度建立最优决策模型；此外，还研究了降低救援成本的风险资产可交易模型，融合巨灾保险研究其风险分散和风险转移的策略；设计相关智能算法对模型进行求解. 为了发掘理论潜在的应用价值，将所得到的研究成果应用到最优检查站点设计、应急物流等实际问题中，为决策者提供决策支持. 本项目主要研究突发事件应急决策中风险分析和动态决策问题，不仅具有重要的学术意义，而且具有重要的实用价值和广泛的应用前景.