基于宏观基本图的定制公交行程时间估测与递进调度方法

Customized bus is a novel developing direction of public transit with personalized and high-quality service. And it also is a representation of strengthening Internet plus transportation services during the “13th Five-Year Plan” period. The core technical problem of customized bus operation is to accurately grasp the traffic conditions of heterogeneous traffic network and the vehicles’ travel time, then to plan routes and dispatch vehicles to respond to heterogeneous passenger requirements. The speeds of customized buses and the macro-traffic flow parameters are estimated by fusing taxi and bus GPS data. With the macroscopic fundamental diagram (MFD) partition theory, the road network is divided into multiple homogeneous and moderate-sized regions which are suitable for customized bus modes. The number of vehicles in the regions are forecasted based on the MFD regulation, and then the travel times of customized buses are forecasted. Aimed to complex travel demands, a progressive scheduling method of “regional path first-actual line second” is proposed. At the meso-level, the vehicles’ regional route planning model is proposed according to regional average travel times. At the micro-level, the actual line planning and vehicle scheduling model is established, which is guided by the regional route and has the objectives of optimizing operating costs, passenger travel costs and demand coverage. This method helps to complete the scheduling decision progressively. This research is not only the exploration of the basic regulation of customized bus, but also the new extension of MFD theory applied in the public transit. It has important practical significance for advancing the theory research and practice in customized bus scheduling.

定制公交是个性化、高品质公交服务的新方向，也将是“十三五”期间强化互联网+交通服务的体现。定制公交运营的核心技术问题是准确把握非均匀路网交通状态及车辆行程时间，据此规划线路并调度车辆响应空间异质需求。通过融合出租车及公交车GPS数据进行定制公交速度及宏观交通流参数估测，利用宏观基本图（MFD）分区理论将路网划分为多个适于定制公交模式且规模适度的均匀子区；基于MFD规律预测子区车辆数进而预测定制公交行程时间；针对复杂的出行需求，提出“子区路径—实际线路”的递进调度思路，在中观层面上，依据子区平均行程时间建立子区级车辆路径规划模型；在微观层次上，以运营成本、乘客出行成本及需求覆盖率最优为目标，建立在子区路径引导下的实际线路规划及车辆调度模型，递进地实现定制公交调度决策。本研究既是对定制公交基本规律的探索，也是MFD理论应用于公交领域的新拓展，对推进定制公交调度理论研究及实践具有重要现实意义。

多元化是城市公共交通的主流发展方向，定制公交则是多元化公交服务中的一种创新模式。定制公交车辆行驶状态受路网交通状态影响，把握影响定制公交的路网交通规律并基于该规律进行定制公交调度是提高决策合理性的重要方式。本项目的主要研究目标是建立适于定制公交运营的子区MFD图谱，基于子区MFD规律估测定制公交行程时间，融合路况进行定制公交调度。.本项目的研究工作主要包括：（1）融合多源交通数据进行公交车流参数估计，采用神经网络模型对缺乏公交运营数据的道路进行速度补齐，设计了公交路网分区算法。该算法采用密度峰值聚类和区域生长法生成初始小区、然后通过小区合并和边界调整得到公交路网子区。与基于K-means的分区算法相比，该方法可得到更合理的公交路网分区。（2）提取公交路网子区流密速参数，获得公交MFD图形。与同时采用社会车和公交车流进行的多模式交通路网分区相比，该方式的公交MFD具有更高的平均拟合优度，可更好地描述公交车流规律；基于公交路网子区速度分布规律，实现定制公交行程时间估测。（3）考虑定制公交速度随机性，建立以服务稳定性和利润最大化为目标的定制公交线网优化模型；依据速度规律估测定制公交行程时间并引导备选线路生成，通过随机模拟仿真寻找最优定制公交线网。实证研究表明，与基于平均速度进行的线网优化相比，该方法可显著提高服务稳定性和利润。（4）针对实时响应式定制公交中，企业在决策时无法获知全部出行需求所造成的困难，提出递进调度方法，首先,基于高概率出行站点规划初始线网，然后,依据初始线路并结合实际出行需求调整路线、调度车辆满足实时出行需求。实证研究表明，设置适宜的线路长度和绕行率有利于减小车队规模；采用软时间窗附加最大延误限制的策略可更好地平衡企业利益和服务水平，且采用全站点控制车辆行驶方向可使服务乘客数量和稳定性更优。.本项目研究成果可为定制公交调度方法提供理论支持，具有重要科学意义。