Introducing lanes and lane changing in analytic dynamic modelling of congested road traffic

在设计、评估、定价和管理道路交通以及拟议的基于信息学、路线指导和其他交通控制措施的智能交通系统时，需要对网络上的道路交通及其随时间的变化(动态)进行建模。路网上的交通流模型可以方便地大致分为两类，即微观模拟(微观模拟)模型和通常被称为分析模型或宏观模型。前者倾向于将交通建模为单独的车辆，而后者将其建模为流。这两种方法都有重要的作用，都有重要的优点和缺点。[例如，微模拟模型可以提供更多细节，而分析模型考虑更广泛的目标和动态，需要更少的数据和参数，运行更快，并具有其他所需的特性。]这两种方法之间的一个显著区别是，在微观模拟中，车道和换道行为的建模起着至关重要的作用。相比之下，在分析模型的当前技术状态中，车道和车道改变在很大程度上被忽略，并且道路通常被视为在每个方向上只有一个车道。这种对比，或者说分析网络模型缺乏现实性，足以解释为什么微模拟模型在实践中越来越多地使用，而不是分析模型，尽管后者及其最近的许多理论和计算进展是必要的。研究的目的是促进和推进将车道和换道行为引入到宏观解析模型中，在尽可能保留解析网络模型现有优势的同时，为这些模型带来微观仿真模型带来的好处和优势。交通从一条车道转换到另一条车道的主要原因之一是为了避免在其他车道上出现拥堵、排队或溢出，因此我们将同时对此类拥堵、排队和溢出的形成和行为进行建模。我们建议以两种不同的方式发展这些模型(包括车道、变道、拥堵、排队和溢出)：第一，通过对现有的用于动态交通分配的各种交通路段模型进行重新表述，以包括车道和变道；第二，通过将车道和变道引入通常的连续单车道交通流理论(流体流和运动波模型)。我们还计划开发每个结果模型的求解算法，并调查和比较每个模型和算法的各种属性、优缺点以及它们在真实性、简单性和计算成本之间的隐含权衡。为了开展上述研究，我们要求为一名研究助理(研究员)提供为期3年的经费，并为计算机设备和参加研究会议的旅费提供一些支助。

Modelling of road traffic and its variation over time (dynamics) on networks is needed for designing, evaluating, pricing and managing road traffic, and for proposed intelligent transport systems based on informatics, route guidance, and other traffic control measures. Models of traffic flows on road network can be conveniently divided into approximately two categories, namely microsimulation (microscopic simulation) models and what are often referred to as analytic models or macroscopic models. The former tends to model traffic as individual vehicles while the latter models it as flows. Both approaches have important roles and both have important advantages and some disadvantages. [E.g., microsimulation models can provide more detail while analytic models consider broader goals and dynamics, require less data and fewer parameters, run faster and have other desirable properties.] A remarkable difference between the two approaches is that, in microsimulation, the modelling of lanes and lane-changing behaviour plays an essential role. In contrast, in the current state of the art for analytic models, lanes and lane changing are largely ignored and roads are commonly treated as having only a single lane in each direction. This contrast, or lack of realism in analytic network models, is sufficient to explain why microsimulation models are increasingly used in practice in contrast to analytic models despite the need for the latter and their many recent theoretical and computational advances. The purpose of the proposed research is to facilitate and advance the introduction of lanes and lane-changing behaviour into analytic macroscopic models, to bring to these models the benefits and advantages that it brought to microsimulation models, while so far as possible retaining existing advantages of analytic network models. One of the main reasons why traffic changes from one lane to another is to avoid congestion, queues or spill-backs in the other lanes, hence we will simultaneously model the formation and behaviour of such congestion, queues and spillbacks. We propose to develop these models (with lanes, lane-changing, congestion, queues and spillback) in two different ways: firstly by reformulating each of various existing traffic link models used in dynamic traffic assignment, so as to include lanes and lane changing, and secondly by introducing lanes and lane changing into the usual continuous single-lane traffic flow theory (fluid flow and kinematic wave models). We also plan to develop solution algorithms each of the resulting models, and to investigate and compare the various properties, advantages and disadvantages of each of the models and algorithms, and the implicit trade-offs they make between, for example, realism, simplicity and computational cost. To undertake the above research we are requesting funding for one research assistant (Research Fellow) for 3 years together with some support for computing equipment and travel to research meetings and conferences.