基于参数诱导的随机动态车队调度求解策略

Solving strategies for the stochastic dynamic fleet scheduling problem based on leading of parameters

基于运输任务产生的随机特性,进行运输任务分布生成机制设计,并给出路网上运输任务分布函数的确定方法,进而研究车辆分布更新过程.在对两大状态变量的更新环节进行设计的基础上,引入车辆增量收益参数,并给出车辆增量收益参数的确定方法及动态更新过程,进而将该参数引入问题模型中的未来时段期望收益函数部分,实现问题模型的可分解形式改造.给出随机动态问题的完整求解算法流程.最后演示独立节点处收益函数拟合过程,并对控制参数对问题求解质量的影响进行数值分析,从而验证方法的有效性.

The number of the transportation tasks between all regions is treated as being random. The gen- erating mechanism on the distribution of transportation task is devised; the formulation of transportation task distribution function is presented; then the transformation mechanism on number of vehicles of a region at a time period to be dispatched is discussed. After the two state variables are updated to improve the solution, the control variable depicting the marginal cost of having one more vehicle in a region is drawn. The calcu- lation and development of the control variable are expressed. Then, using the control variable to approximate the expected function in the model, the dissolution model is formulated. The algorithm is given. At last, the revenue function＇s fitting process at a single point is simulated. The effect of the control variable in improving the solution quality is numerically analyzed. It is verified that the control variable is very helpful to enhance the solution quality.