【作者】 于滨；

【导师】 程春田； 杨忠振；

【作者基本信息】 大连理工大学 ， 计算机应用技术， 2006， 博士

【摘要】 优先发展城市公共交通系统是解决大、中城市交通问题的最佳途径。近年来，城市公交系统的智能化已经成为公共交通领域研究的主要方向。本文以大连市公交系统为工程背景，研究了蚁群算法、SCE-UA算法以及支持向量机等智能算法在公共交通规划和调度中的应用。城市公交系统是一个巨系统，其相关的模型和方法都非常复杂，为此，本文引入了高性能计算来提高智能算法的优化质量和收敛速度。本文的主要内容如下： 1)公交线路网是城市公共交通系统的根本和基础，其设置得是否合理会直接影响乘客的出行时间、换乘以及系统运营成本。本文提出了以直达客流密度最大为目标的公交线网优化模型。该模型综合考虑了乘客和运营者双方的利益，通过一种模拟进化算法——蚁群算法，来优化该模型。为了提高蚁群算法的搜索效率，采用了综合考虑全局和局部信息的信息素增量更新策略：ANT-Weight策略和基于粗粒度模型的并行方案。以大连市主城区的公交数据对该模型和算法进行了检验，与大连市现状的公交网络相比，基于本文模型优化的公交网络的效率更高。另外，结果还表明ANT-Weight策略和粗粒度模型可以提高蚁群算法的效率。 2)发车频率的制定是公交系统日常运营工作的核心，它决定了运行时刻表、车辆调度以及分派司机等其它的日常调度工作。本文提出了一个双层规划模型来优化公交线路发车频率，该模型可以反映运营者(供给者)和出行者(需求者)之间的相互作用。其中，上层模型以整个公交系统总成本最小为目标，通过进化算法——SCE-UA算法来优化公交线路的发车频率；下层模型基于最优出行策略，进行公交客流分配。以大连市主城区的公交系统为计算实例分析了该双层模型及其解法的效率，结果表明，该双层模型可以有效地节省系统的总成本。 3)开发能准确预测公交车辆到站时间的方法，是提高公交车辆准时性，减少乘客等待时间，从而提高公交吸引力的重要手段。本文提出了一种基于支持向量机(SVM)和Kalman滤波的混合模型，来预测公交车辆到站时间。在该混合模型中，SVM模型基于历史数据预测路段的基线运行时间；基于Kalman滤波的动态算法结合基线时间和最新的车辆运行信息，预测车辆到达各站点的时间。以大连市开发区7路公交线路的数据对该方法进行了检验，实例分析表明，与SVM模型相比，该混合模型的稳定性和预测精度都较高。 4)实时调度策略就是利用先进的技术手段，动态地获取实时交通信息，实现对车辆的实时监控和调度。本文主要研究了实时调度中最常用的滞站调度策略，提出了两种滞更多还原

【Abstract】 Preferential development of urban public transportation system is a universally accepted approach of solving urban traffic problem. Recently, intelligent public transportation system has become the important development trend of public transportation field. Based on the project background of public transportation system in Dalian city, the applications on ant colony optimization algorithm, support vector machine, SCE-UA algorithm, etc in the planning and the operation of urban public transportation system, are studied. Since urban public transportation system is a giant system, the corresponding models and strategies are very complex. In this dissertation, the parallel intelligent algorithms running in PC cluster are adoped to improve the optimization quality and speed. The major contents and research progress are as follows:1) Bus network is the basis of urban public transportation. The rationality of the bus network, therefore, directly influences the travel time and transfer rate of passengers, and the overall running cost of the transport systems. This dissertation presents an optimization model for bus network design, which aims to maximize the number of direct travelers per unit length, i.e. direct traveler density, subject to route length and non-linear rate constraints. Ant colony optimization (ACO), a new evolution algorithm, is used to solve the model. To improve the efficiency of the method, two improved strategies are proposed: 1) develops a new strategy to update the increased pheromone, called Ant-Weight, which considering the global and local information; 2) uses parallelization strategies of ACO improve the calculation time and the quality of the optimization. The data of Dalian city in China are used to test the model and the algorithm. The results show that compared with the preseng transit network the optimized one is more effective and efficient. They also reveal that the ANT-Weight strategy and coarse-grain strategy are effective.2) The optimization of the frequencies is the key of the operation plan, which determine the situation of the running schedule, vehicle adjustment and driver assignment. Considering the interactions between the operators (the demand) and the user (the supply), in this dissertation, a bi-level programming model for the bus frequencies design in the given network is presented. In the bi-level model, the upper-level is the leader and optimizes the bus frequencies by SCE-UA. The lower-level is the follower and its objective is to assign transit trips to the network based on travel strategy and the optimal frequencies. Finally the model and the algorithms are illustrated with the bus network in the city of Dalian in China. The results show that the model can effectively save the total cost of the operators and users更多还原