【作者】 杨英俊；

【导师】 赵祥模；

【作者基本信息】 长安大学 ， 智能交通与信息系统工程， 2013， 博士

【摘要】 国外大多城市和地区人口密度较小，出租汽车行业市场化程度较高。我国由于人口众多、城市处于高速发展等原因，出租汽车行业完全市场化将使得竞争无序，并且服务质量难以保证，因此，把出租汽车定位为公共交通的补充，我国绝大多数城市出租汽车行业的管理都采用政府管制为主的模式，但由于出租汽车以“扬召”为主的营运模式，政府管制也存在信息来源受限、决策困难、决策缺乏依据、利益难以平衡等弊端。针对以上问题，论文在对现阶段的出租汽车行业的管理模式、存在问题以及国内外的相关研究深入分析的基础上，结合我国城市出租汽车营运和管理的特点，对以车载数据采集终端为核心的出租汽车营运信息采集模式和方法进行了研究。基于多个城市的采集数据，论文对出租汽车投放量、出租汽车定价和电召中心值守班次等出租汽车决策和管理中的关键问题进行了研究。最后对出租汽车服务管理系统进行了设计。主要创新点包括：（1）提出一种基于出租汽车实时营运数据的精细预测和中长期预测结合的出租汽车运力投放模型。在对现阶段出租汽车投放存在问题进行分析的基础上，采用精细预测与中长期预测相结合的方式进行预测。精细预测采用城市基础参数、历史数据和实时出租汽车营运数据作为预测数据，在出租汽车平均工作时间、城市居民平均收入等条件约束下建立预测模型。中长期预测采用国民生产总值、市区人口、居民消费水平、平均候车时间、公交线路总长等出租汽车影响因素的历年数据作为输入，以小波神经网络作为预测工具，对城市中长期出租汽车保有量进行预测。（2）提出一种基于特征价格模型理论的城市出租汽车定价方法。该方法采用经济学中的特征价格模型理论对出租汽车的定价问题进行研究，得出适合出租汽车定价特征的价格模型，该模型以车型、人均可支配收入、公交系统发展水平、出租汽车空驶率、城市出租汽车投放量、市区常住人口数、交通拥挤程度、旅游业发展水平等为特征，采用多元线性回归的方法进行定价的计算。（3）提出一种基于遗传算法的出租汽车电召人员的值守班次确定方法。电召方式能够为乘客提供预约式服务，为乘客和司机都带来收益，电召中的值守人员成本是主要成本之一。论文以分时段的话务量为基础数据，采用遗传算法对值守人员的班次进行确定。该方法解决了手工排班计算复杂的弊端，能够以较快的方式获得最优的排班方案。论文所提出模型和开发的系统在全国30个城市的出租汽车服务管理试点工程中进行了应用，应用结果表明，论文工作能够为我国出租汽车的实际管理工作和决策支持提供方法，具有较高的理论和应用价值。更多还原

【Abstract】 In general, the level of marketization of taxi industry is high in other country where thepopulation density is low. China has a large population density and is experiencinghigh-speed development and urbanization. Such unique characteristics indicate that the taxiindustry in China cannot be completely dependent on the market. Otherwise, the marketcompetition can end up in chaos, resulting in the loss of level of service. Therefore, the taxiservices in most cities in China are currently managed mainly by the government andregarded as a supplement for the public transportation. Nevertheless, the taxi services inChina are operated primarily based on a “hailing” mode, which has inherent drawbacks formanagement. For example, there is a lack of source of information; the decisions are not wellsupported by sound references; and it is difficult for the government to make reasonabledecisions and balance benefits of different stakeholders.Regarding the problems mentioned above, this dissertation presents a research thatstudied a method to collect the information of taxi operations using intelligent on-vehicledevices, based on the analysis of existing management modes, problems and related nationaland international researches in the taxi industry. This research also investigated some keyproblems related to the decision and management activities in the taxi industry, including thequantity of taxis in a market, the pricing of taxi services and the staff scheduling strategies ofthe taxi call centers. Further, this research introduced a new design of a management systemfor taxi services. The innovation of this work can be elaborated with the following aspects:(1) A prediction model to determine the quantity of taxis in a market was developedbased on the real-time data of taxi operations. This model combined precise and mid-to-longterm prediction and was based on a full analysis of existing problems regarding thedetermination of the quantity of taxis in a market. The precise prediction model used somecharacteristics of a city, including the basic urban parameters, the historical data and thereal-time data of taxi operations, as model inputs and was developed under the constraint ofaverage time of taxi in operation and average citizen income, etc. The mid-to-long termprediction model used some characteristics of a city, including the gross domestic product, thepopulation, the household consumption level, the average waiting time for taxis and the totallength of public transport lines, etc in each of many historical years, as model inputs and adopted the wavelet neutral network as a tool to predict the mid-to-long term quantity of taxisin the market of this city.(2) An urban taxi pricing method was developed based on the hedonic price theory. Thismodel used vehicle type and some characteristics of a city, including the per capita disposableincome, the development level of public transport systems, the ratio of the taxis in idlingstates, the quantity of taxis in the market, the total number of permanent residents, the level oftraffic congestion and the development level of tourism, etc as the model inputs anddetermined the pricing strategy using a multiple linear regression.(3) A staff scheduling method for taxi call centers was introduced based on a genericalgorithm. This method used the call center’s workload in different time-of-day as one of thebasic input data and applied a generic algorithm to determine the staff scheduling. Anoptimized schedule could be achieved in a fast fashion, overcoming the disadvantages of thetraditional manual scheduling method, reducing the cost of staff in call centers and bringingbenefits to both customers and drivers.The models and systems developed in this research described by this dissertation hadbeen showcased in30pioneer cities in China. The result showed that this work could providescientific methods to guide the management and decision activities in the taxi industry inChina and had a relatively high value in both theory and application.更多还原