【作者】 诸葛承祥；

【导师】 邵春福；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2014， 博士

【摘要】 国民经济的快速增长加快了城市化的进程,城市规模不断扩张,机动化进程加快,居民出行需求猛增。城市交通为社会发展做出巨大贡献的同时,也因出行需求增长过快诱发了许多问题,城市道路交通拥堵问题已经成为困扰世界各国政府和交通研究者的难题之一。城市土地利用和交通是“源”和“流”的关系,国内外实践证明,二者的协调发展是从根源上解决城市交通问题、实现土地利用集约化的重要前提,是决定城市能否实现可持续发展的关键。本论文以城市交通和土地利用(UT-LU)为研究对象,利用自组织理论对其宏观和微观演化机理进行研究。其中,宏观部分的研究利用宏观自组织理论中的耗散结构理论、协同学理论、超循环理论、突变理论、混沌理论和分形理论,分别对UT-LU的自组织演化条件、演化动力、演化结合形式、演化途径、演化图景和演化结构复杂性进行揭示。并以北京市为例,对宏观自组织理论在UT-LU中的适用性进行验证。微观部分的研究借助多智能体技术,对UT-LU各关键部分的微观自组织原理进行分析,主要包括出行需求、交通网络状态、居住地选择、房产价格等自组织原理的研究,在此基础上,按照先后顺序,建立各关键部分之间的联系,并通过相互调用,实现UT-LU的微观自组织演化。此外,还对UT-LU演化分析所需的微观数据合成方法和仿真系统SelfSim进行研究,具体内容如下：(1)微观数据合成方法研究。仿真人口和初始日活动计划是UT-LU微观演化研究所需的基础数据,因此,本论文首先对人口合成方法和初始日活动计划生成方法进行研究。针对传统人口合成方法未考虑初始家庭权重以及方法的优化目标未考虑控制变量平均误差的离散程度,提出了一种新的启发式算法作为人口合成工具以克服上述两方面的不足。此外,在人口合成的基础上,本论文将效用最大化理论和遗传算法相结合,用于家庭成员的初始日活动计划生成,为活动计划调整提供初始方案。最后,以保定市为例,对人口合成方法和初始日活动计划生成方法中的参数设置进行讨论。(2)基于多智能体的活动安排和交通流分配联合模型。考虑到居民出行需求和交通网络状态存在相互作用,因此在研究JT-LU微观自组织演化机理时将二者进行联合建模。联合模型输入合成的人口和初始日活动计划,以家庭效用最大化为目标,并利用多智能体技术对家庭成员活动安排和交通网络状态之间的动态反馈进行模拟,输出最终的活动计划和交通网络流量状态。具体建模内容包括仿真环境和多智能体定义以及仿真核心模块构建。其中仿真核心模块包括执行计划、计划评分和重新计划三个子模块,通过三个子模块的循环调用,活动安排和交通网络状态最终趋于合理。最后,联合模型在保定市进行了应用,结果表明模型仿真精度能满足应用要求。(3)基于多智能体的居住地选择和房产价格联合模型。在UT-LU微观自组织演化中,居住地选择和房产价格存在互馈关系,因此在分析二者自组织原理的基础上进行联合建模。利用多智能体技术,对家庭智能体的住宅选择行为、业主智能体的价格更新行为以及二者之间的协商行为进行建模,令家庭智能体和业主智能体在协商过程中分别确定居住地和住宅价格。在协商行为建模时引入融合了前景理论的效用最大化理论,用于对家庭智能体的住宅选择行为进行建模,另外还涉及建立基于多智能体微观仿真的可达性模型以用于住宅可达性的计算,为家庭智能体的购房决策提供依据。最后,以保定市的房地产市场为例,对模型的预测精度进行了分析,结果表明模型预测能力较强。(4)人口发展模型。人口发展是推动UT-LU演化的重要动力来源,人口发展模型主要对仿真场景中与UT-LU演化相关的人口状态以及特征演变进行建模,具体包括出生模型、上学模型、就业模型、结婚模型、死亡模型、小汽车拥有量模型和月收入模型。(5) SelfSim仿真系统。对UT-LU各关键部分的演化模型进行整合,搭建SelfSim仿真系统,用于UT-LU微观自组织演化研究。最后,利用SelfSim对2008-2013年保定市UT-LU的短期动态演化过程进行仿真模拟,并将仿真结果和实际调查数据进行比较分析,结果表明SelfSim在多方面具有较高的仿真精度。更多还原

【Abstract】 The rapid development of national economy facilitates the process of urbanization and mechanization, expands the urban scale and increases the travel demand of urban residents. Obviously, the urban transport contributes a lot to the development of society, however, it also generates many problems. Traffic congestion in metropolises is a big headache for governments and transport planners all over the world. The relationship between urban transport and land use can be regarded as the link between "source" and "stream". It has been proved that the coordinate development between land use and urban transport is a prerequisite for resolving the traffic problem fundamentally and realizing the intensification of land use. Meanwhile, it is a key to the sustainable urban development.This paper focuses on "Urban Transport-Land Use"(UT-LU) and conducts a study on the evolution mechanism of the UT-LU from the macro and micro perspectives based on Self-Organization Theory. From the macro perspective, the Dissipative Structure Theory, the Synergetics Theory, the Hypercycle Theory, the Catastrophe Theory, the Chaos Theory and the Fractal Theory will be used to reveal the UT-LU’s evolution condition, evolution power, evolution binding form, evolution approach evolution prospect, evolution Structural Complexity, respectively. And then the paper takes Beijing as an example to examine the applicability of the macro perspective of Self-Organization Theory in the UT-LU.From the micro perspective, the Multi-Agent technology is adopted to analyze the micro self-organizing evolution mechanism of each component, such as travel demand, transport network status, residential location choice, real estate price, etc., in the UT-LU. Then establish the relationship among these components according to sequencing. By mutually calling these components, the self-organizing evolution of UT-LU can be implemented. In addition, the micro-data synthesis method and simulation platform SelfSim, which are used for analyzing the evolution of UT-LU, are also studied. The details about the above study are as follows:(1) Micro-data Synthesis Method. Both population and initial plans are the basic data required by the study on micro-evolution of UT-LU. Since the traditional population synthesis method doesn’t take the initial household weights into consideration and its optimization objective ignores the discrete extent of average error of control variables, a new heuristic algorithm is proposed to overcome the limitations mentioned above. In addition, the approach of generating initial plan is employed to generate an initial plan for each agent in SelfSim. And the Utility Maximization Theory and Genetic Algorithm are integrated into the approach. The initial plan will be treated as an initial solution to the plan adjustment. Finally, take Baoding as an example, the parameters setting of population synthesis method and the approach of generating initial plan is discussed.(2) Multi-Agent-Based Coupled Model of Activity Schedule and Traffic Assignment. Since the interaction between travel demand and transport network status, both of them are modeled together in the study of micro self-organizing evolution of UT-LU. After the synthetic population and initial plan are both input into the coupled model, the final plan of each agent and transport network status will be output. Actually, the Multi-Agent technology here is used as a tool to simulate the interaction between travel demand and transport network status. The modeling contains defining the agent and simulation environment, as well as establishing the core part of simulation. As for the core part of simulation, it involves the sub-modules of execution, scoring and reschedule. These three sub-modules form a cycle in calling, and then the final plan and transport network status are output after certain iterations. Finally, the coupled model is applied to Baoding and the results show that the forecasting accuracy of the model meets the requirements of application.(3) Multi-Agent-Based Coupled Model of Residential Location Choice and Real Estate Price. Since the interaction between residential location choice and real estate price is in UT-LU evolution, residential location choice and real estate price are modeled together. The Multi-Agent technology is employed to model the Household Agents’residential selection behavior, Owner Agent’s behavior of updating price and the communication behavior between Household Agent and Owner Agent. The Household Agents’selection of their houses and the Owner Agents’pricing of their houses will finalize through their communication. Here the Utility Maximization Theory and the Prospect Theory are both employed to model the residential choice behavior of the Household Agents. Meanwhile, the Multi-Agent-Based Micro-Simulation Accessibility Model is developed to calculate the house accessibility, which is an influencing factor in Household Agents’selection of houses. Here, the coupled model is also applied to Baoding and the forecasting results demonstrate that the model gains a satisfactory forecasting ability. (4) Population Development Model. The population development is one of the crucial factors for promoting the evolution of UT-LU. The Population Development Model attempts to model the status and features of the population which are closely related to UT-LU evolution. This model mainly includes Birth Model, Education Model, Work Model, Marry Model, Death Model, Car Ownership Model, and Income Model.(5) Simulation Platform of SelfSim. By integrating all the models of key component mentioned above, the SelfSim is developed for the analysis of UT-LU evolution. Finally, the short-term evolution of UT-LU of Baoding from2008to2013is simulated and its results show that SelfSim has the high forecasting ability in a few sides.更多还原