Highway Network Digital Twin for Traffic Management

公路网效率低下降低了通行能力，增加了用户和货物的行程时间。道路通行能力在空间或时间上的使用并不均衡：早晚高峰时段的使用高峰导致效率低下;事故意外减少了通行能力。实时预测交通模式的能力与交通控制系统相关联，可以提高网络效率。基于计算机的交通模型可以计算个人所走的路线，使我们能够在交通网络中复制单个时间点的真实情况。然而，这些模型并不适合应对不可预见的事件，这需要快速的运营决策来减轻其影响。网络上特定点的交通状况越来越多地被路边传感器捕获，这些传感器实时监测交通量，车辆排放和其他信息。这些数据非常有价值，许多数据经常用于交通管理;然而，它们只告诉我们空间和时间中特定点的情况，而不是整个网络的情况。使用机器学习等人工智能技术处理和识别此类数据中的模式的技术正在改进，并提供将这些数据应用于更复杂的实时应用的机会。如果有一个可靠的，准确地表示整个现有道路网络及其在现实世界中的运行状况--一个“数字孪生”。当人们选择的路线我们的建议是基于现有的交通模型、不断增长的路边传感器阵列和最先进的机器学习技术建立一个实时数字孪生模型，以便能够实时模拟网络上的交通状况。数字孪生模型将允许主动管理交通，增加道路容量和促进更顺畅的交通流量，特别是在“高峰时间”和意外事故期间。例如，在实施这些措施之前，数字孪生模型可以用来预测改变标志或打开/关闭车道对交通流量（以及容量）的影响。为了开发和运行实时数字孪生模型来管理和增加道路容量，我们必须首先证明我们可以自动集成实时传感器数据、道路网络表示和交通模型。我们将评估这种集成水平的可行性，沿着运行实时解决方案在计算速度方面的实用性。

Public descriptionInefficiencies in road networks reduce capacity and increase journey times for users and goods. Road capacity is not used evenly over space or time: usage peaks during morning and evening rush hours lead to inefficiencies; incidents reduce capacities unexpectedly. The ability to forecast traffic patterns in real-time, linked to traffic control systems, can improve network efficiency.Computer-based traffic models that calculate routes that individuals take allow us to replicate real-world conditions across a traffic network for a single point in time. However, these models are not well-suited to responding to unforeseen incidents, which require rapid operational decision-making to mitigate their impacts.Traffic conditions at specific points on the network are increasingly being captured by roadside sensors that monitor traffic volumes, vehicle emissions and other information in real-time. These data are very valuable, and many are routinely used for traffic management; however, they tell us only about conditions at specific points in space and time, and not conditions across the entire network.Techniques for processing and identifying patterns in such data, using Artificial Intelligence techniques like Machine Learning are improving, and offer opportunities to apply these data in ever more complex real-time applications.Traffic management can be significantly improved if there is a reliable, accurate representation of the whole of the existing road network and the real-world conditions it is operating under -- a 'digital twin'. Such a digital twin becomes a powerful forecasting tool when the routes that individuals take (or should take) are accounted for.Our proposal is to build a real-time digital twin based on existing traffic models, this growing array of roadside sensors, and state-of-the-art Machine Learning techniques in order to be able to model traffic conditions on networks in real-time.The digital twin will allow traffic to be actively managed, increasing road capacity and facilitating smoother traffic flow, particularly during "rush hour" and during unanticipated incidents. A digital twin could, for example, be used to predict the effects on traffic flow (and therefore capacity) of changing signs or opening/closing lanes before implementing such measures. In order to develop and operate a real-time digital twin to manage and increase road capacity, we must first demonstrate that we can automatically integrate real-time sensor data, a road network representation and traffic model. We will assess the feasibility of this level of integration along with the practicality of running a real-time solution in terms of calculation speed.