Development of New Generation Microsimulation Models for Transport Demand

在过去的20年里，我的大部分研究都是在开发计算机模型来模拟城市地区的土地使用和交通过程。土地利用过程是指我们在空间上组织生活的方式，即我们居住、工作、购物和从事其他活动类型的地方。交通过程是指个人参与此类活动的需要所产生的交通需求/旅行方式。实证研究表明，交通需求与土地利用之间存在双向关系。我开发的模拟模型，被称为综合运输和土地利用模型（ITLUM），试图捕捉这种关系。这些模型可以被规划者用作决策支持工具，因为它们允许对政策的潜在后果进行调查。虽然ITLUM领域已经发展了40多年，但在20世纪90年代出现了从聚合模型到分解模型的明显转变。在前者中，城市被概念化为构成可观察或分析单元的区域的集合。根据这些区域的特点，建立了一组方程来估计区域之间产生的行程以及这些行程在交通网络中的分配。相反，分解模型将城市视为个人、家庭和企业（统称为“代理人”）的集合，这些人对自己的日常活动做出决定。这些决定往往导致旅行。虽然向自下而上或基于主体的模型的转变试图捕捉主体的行为，但仍有三个方面阻碍了该领域的进展。首先，主体的行为通常是在独立于决策环境的情况下进行研究的。第二，这个环境是动态的，决策必须适应它。第三，缺乏一个被广泛接受和灵活的开源平台，在这个平台上，来自世界各地和各种学科的研究团队可以做出贡献。目前，这些努力是支离破碎的。我认为，如果要取得重大进展，所需要的无非是范式的转变。在过去的三年里，我开始研究复杂自适应系统（CAS），这是一个在过去20年里经历了爆炸式增长的领域，但在ITLUM中却没有。CAS是指由自适应主体群体组成的自然和人工系统，它们的相互作用导致可能与现实世界有一些相似之处的紧急行为。找到一组与这些代理相关的规则，将有助于模拟现实世界的行为，并帮助我们预测、控制或解释复杂的系统。我认为，在国际贸易法委员会背景下对CAS进行研究有可能克服上述三个障碍。因此，本研究的总体目标是为ITLUM开发CAS框架，并将其性能与先前的模型进行比较。为此，我们的子目标是：1。对有关CAS的文献以及可用于开发此类系统的开放软件平台进行全面的回顾。2. 在一个地理信息系统平台上为2001年汉密尔顿人口普查都市区的家庭、个人和企业编制综合人口数据，作为试验台。3. 使用CAS和合成种群开发最终ITLUM的组件，并使用它进行广泛的实验。4. 使用CAS为Hamilton设计并实现ITLUM。5. 与更传统的模型相比，彻底评估已开发系统的性能。作为这项研究的一部分，开发一个开放平台，并致力于在ITLUM上下文中使用CAS，为该领域提供了一场革命的可能性。

Over the last 20 years, a large part of my research has been in the development of computer models that simulate land use and transportation processes in urban areas. Land use processes refer to the way we organize our lives over space, that is, where we live, work, shop and pursue other activity types. Transportation processes refer to the way transport demand/travel is generated by the needs of individuals to participate in such activities. Empirical research has demonstrated that a bi-directional relationship exists between transport demand and land use. The simulation models I developed, known as Integrated Transport and Land Use Models (ITLUM), attempt to capture this relationship. Such models can be used as decision support tools by planners for they allow the investigation of potential consequences of policies. While the field of ITLUM has evolved over four decades, in the 1990s there was a noticeable shift from aggregate to disaggregate models. In the former, a city is conceptualized as an aggregation of zones that constitute the observables or units of analysis. Based on traits of these zones, a set of equations estimate trips that are generated between zones and the allocation of those trips on the transportation network. In contrast, disaggregate models view a city as a collection of individuals, households and firms (collectively "agents") that make decisions on their daily activities. These decisions often result in travel. While the shift towards bottom-up or agent-based models attempts to capture the behaviour of agents, there remain three aspects that hamper progress in the field. The first is that the behavior of agents is often studied in isolation from the environment within which decisions are made. The second is that this environment is dynamic and decision-making must adapt to it. Third is the lack of a widely acceptable and flexible open-source platform within which research teams from around the world and from a variety of disciplines can contribute. At the moment, efforts are fragmented and disconnected. I contend that if significant progress is to be made, what is required is nothing short of a paradigm shift. In the last three years, I started work with Complex Adaptive Systems (CAS), a field that witnessed an explosion over the last 20 years but not in ITLUM. CAS refers to natural and artificial systems composed of populations of adaptive agents whose interactions lead to emergent behaviour that may bear some resemblance to the real-world. Finding a set of rules that are relevant to these agents, will help to simulate real-world behaviour and help us predict, control, or explain complex systems. I believe that research into CAS in the ITLUM context has the potential to overcome the three obstacles mentioned above. Accordingly, the overall objective of this research is to develop a CAS framework for an ITLUM and to compare its performance against prior models. To this end, our sub-objectives are to: 1. Perform a thorough review of the literature on CAS along with the open software platforms that can be used to develop such systems. 2. Develop a synthetic population for the Hamilton Census Metropolitan Area at the household, individual and firm level for 2001 in a Geographical Information System platform to be used as a test bed. 3. Develop the components of a final ITLUM using CAS and the synthetic population, and perform extensive experiments with it. 4. Design and implement an ITLUM for Hamilton using CAS. 5. Thoroughly evaluate the performance of the developed system vis-a-vis more conventional models. The development of an open platform, as part of this research, and dedicated to the use of CAS in the ITLUM context, offers the possibility to be nothing short of a revolution in the field.