CARS : Connecting Administrative vehicle data for Research on Sustainable transport

我们的提案建立在之前UKRI资助的项目（称为“EPSRC MOT项目”）的基础上，是一个及时和令人兴奋的机会，可以链接和开放与英国（GB）所有小于3.5吨的车辆（“轻型车辆”）的注册和使用模式相关的两个数据集：1。“注册数据”：DVLA的车辆许可程序生成车辆库存表，每个季度为大约4000万辆车辆编制，记录车辆使用寿命期间注册管理员的位置以及许多车辆属性。运输部（DfT）每季度收到这一数据，并公布汇总的车辆许可统计数据，其中车辆特征数量非常有限。车辆查询服务API可通过特别许可访问，但可用字段有限。2.“MOT数据”：自2006年以来，DVSA已经以数字方式记录了每辆在GB中进行年度道路行驶性能（“MOT”）测试的车辆的品牌、型号、年龄、发动机类型和里程表阅读。每年增加约3000万次测试和5000万个相关项目。这些数据自2010年底以来一直公开发布，但数据库只包括首次测试的车辆（通常是三年后），因此它只是所有车辆的部分记录。公开释放的数据也不包括登记饲养人的所在地。根据MOT数据中提供的里程表阅读为每辆车创建年里程数在计算上是复杂的，并且形成了我们先前研究中进行的工作的核心部分（参见我们的初步研究的详细信息，我们将基于该方法）。目前，除了数据所有者之外，任何人都不可能将这些数据源链接起来，以创建一个完整的所有车辆和里程表读数的数据库，并将这些数据与它们在该国的位置联系起来。这是因为目前只有来自这两个来源的汇总或部分数据公开发布。此外，即使在MOT数据中提供里程表读数的情况下，年度车辆里程的创建在计算上也是复杂的，并且需要大量的努力和资源来产生稳健的数据。这些计算构成了我们以前研究中所做工作的核心部分，我们将以这些创新为基础制定本项目的方法。目标是将这两组由驾车公众生成的数据常规地结合起来，里程计算标准化和自动化，并提供匿名的单组数据。我们希望从这个项目中产生多个研究就绪的数据集。我们的核心“最小可行产品”（MVP）将包括：1。“车辆级数据”是最具扩展性的，存储在可信研究环境（TRE）中（并从中派生出其他产品数据集）。2.“聚合车辆数据”，其聚合到适当的地理位置，例如LSOA，并且也存储在TRE内。我们将与数据所有者合作，简化驾车公众生成的这两组数据的常规汇总和匿名方式，以便无法识别单个车辆。我们将开发各种方法来增加每辆车的里程和排放量，并进行一些探索性的数据，以展示如何使用这些数据来调查与环境影响，道路安全，可达性以及交通和能源基础设施的有效和公平投资有关的一系列社会目标。

Our proposal builds on a previous UKRI funded project (referred to as the 'EPSRC MOT Project') and is a timely and exciting opportunity to link and open up two data-sets related to the registration and usage patterns of all vehicles less than 3.5 tonnes ('Light Duty Vehicles') in Great Britain (GB): 1. "Registration data": the DVLA's vehicle licensing procedures generate vehicle stock tables, compiled for about 40 million vehicles each quarter, which record the location of the registered keeper through the vehicle's lifetime as well as numerous vehicle attributes. The Department for Transport (DfT) receives this data quarterly and publishes aggregated vehicle licencing statistics with a very limited number of vehicle characteristics. A Vehicle Enquiry Service API can be accessed via special permission but the available fields are limited. 2. "MOT data": since 2006, the DVSA has digitally recorded the make, model, age, engine type and odometer reading of every vehicle taking an annual road worthiness ('MOT') test in GB. Approximately 30 million tests and 50 million associated items are added each year. This data has been on public release since late 2010 but the database only includes a vehicle once it has its first test (usually after three years) and so it is only a partial record of all vehicles. The data on public release also does not include the location of the registered keeper. The creation of annual mileages for each vehicle from the odometer reading supplied in the MOT data is computationally complex and formed a core part of the work undertaken in our previous study (see details of our initial study on which we would base the methodology). It is currently impossible for anyone other than the data owners to link these data sources to create a complete database of all vehicles and odometer readings and link these to where they are located in the country. This is because only aggregated or partial data from these two sources is currently on public release. In addition, even where the odometer readings are supplied in the MOT data, the creation of annual vehicle mileages is computationally complex and requires a significant amount of effort and resource to produce robust data. These computations formed a core part of the work undertaken in our previous study and we would base the methodology in this project on those innovations. The goal is that these two sets of data generated by the motoring public are routinely combined, mileage calculations are standardised and automated and an anonymised single set of data is made available. We expect to produce multiple research-ready datasets from this project. Our core 'minimum viable product' (MVP) will consist of: 1. 'Vehicle-level Data' which is the most disclosive and stored inside a Trusted Research Environment (TRE) (and from which other product datasets are derived). 2. 'Aggregated Vehicle Data' which is aggregated to an appropriate geography, such as LSOA and also stored within a TRE. We will work with the data owners to streamline the way in which these two sets of data generated by the motoring public can be routinely brought together and anonymised so that individual vehicles cannot be identified. We will develop ways of robustly adding mileages and emissions of each vehicle and carry out some exploratory data to show how this data can be used to investigate a whole series of societal objectives related to environmental impacts, road safety, accessibility and efficient and fair investment in transport and energy infrastructure.