即时响应式定制公交调度优化

Optimization of Real-time Responsive Customized Bus Dispatch

调度是支撑即时响应式定制公交运营的关键技术。针对即时响应式定制公交高度分散和随机的乘客出行需求的特点,建立了即时响应式定制公交两阶段调度决策模型。第1阶段进行定制公交初始线路整体决策,以车辆数(线路数)最少为目标,根据区域内分时段的高概率出行OD点的地理分布,优化定制公交系统的初始线路;第2阶段进行车辆实时调度决策,以乘客延误成本最小、运输企业利润最大以及未服务乘客造成的损失最小为目标,在初始线路的基础上,结合实时乘车请求的时空分布、上/下车站点关系、上/下车时间、车辆容量等限制条件,对各线路车辆的实际行驶路线以及到站时刻进行决策。两阶段调度方法从整体和局部两个层面平衡了运输企业和乘客双方的利益,在车辆实时调度决策中兼顾了实时需求和后续最可能需求对调度决策方案的影响。根据两阶段调度模型的特点,分别设计了改进的遗传算法和带精英策略的快速非支配排序遗传算法(NSGA-Ⅱ)。最后,以广州市内的高概率出行点为例对即时响应式定制公交两阶段调度模型和算法进行了验证。仿真结果表明:初始线路优化模型能够生成数量最少且覆盖区域内所有高概率出行点的线路,车辆实时调度决策模型能够根据实际乘车请求合理调整车辆的行驶路线和到站时刻。

Dispatching is a key support technology in the operation of real-time responsive customized bus.Aiming at the characteristics of highly dispersed and random passenger travel demand in real-time responsive customized bus system,we established a two-stage dispatch decision model for the real-time responsive customized buses.In the first stage,we made the overall initial route decision of customized buses.With the minimum number of vehicles(routes)as the goal,we optimized the initial routes of customized bus system according to the geographical distribution of time-phased high probability travel OD points within the area.In the second stage,we made the real-time decision on vehicle dispatching.With the goal of minimizing passengers’delay costs,maximizing profits of transport enterprises,and minimizing losses caused by unserved passengers,based on the initial routes,we made the decisions on the actual driving routes and arrival time of vehicles on each route combining with the temporal and spatial distribution of real-time ride requests,the relationship of on/off stops,on/off time,and vehicle capacities,etc.The two-stage dispatch method balances the interests of transport enterprises and passengers from the overall and individual levels,and considers the effects of real-time demand and subsequent most likely demand on the real-time dispatch decisions.According to the characteristics of the two-stage dispatch model,we designed an improved genetic algorithm and a fast non-dominated sorting genetic algorithm with an elite strategy(NSGA-Ⅱ)respectively.Finally,taking the high-probability travel points in Guangzhou for example,we verified the two-stage dispatch model and the algorithm for real-time responsive customized buses.The simulation result shows that the initial route optimization model can generate the least number of routes that cover all high-probability travel points in the area,and the vehicle real-time dispatch decision model can reasonably adjust the driving route and arrival time of the vehicle according to the actual travel request.