Avoiding Pollution Hotspot in Post COVID-19 Era: An AI-based Air Pollution Exposure (APEX) Visualisation Platform for Individuals and UK Regions

现有的导航应用程序(如Google Maps、Apple Maps、Waze)只能提供实时交通信息，尽管长期暴露在**空气污染(即无机和有机污染物)中是对人类健康的最大环境威胁(Public Health England，2020)**。高浓度的无机空气污染物(如PM2.5、PM10、二氧化碳、NO、NO2等)与破坏性健康疾病(如中风、肺癌、哮喘)有关(世卫组织，2016)，而有机空气污染物，特别是气雾剂，最近已被认为是生物制剂(如冠状病毒)的活跃运输者(Wang&Du，2020)。因此，迫切需要找到创新和可持续的方法来监测空气污染物并遏制其对民众的破坏性影响(DEFRA，2020)。尽管多年来这种紧迫感如此迫切，但有证据表明，**“缺乏可持续和通用的空气质量监测仪器/系统”**是目前没有公开可用的系统/平台让英国人安全地导航污染热点及其毁灭性影响的主要原因。尽管存在参考空气质量监测仪器(例如Highways England的空气质量分析仪)，但众所周知，它们**安装和维护**非常昂贵**；因此，这种参考仪器不能相对地部署在整个英国以捕获污染物浓度水平。此外，最近流行的低成本传感器(如Zephr、AirScan)已知存在**精度、可靠性和电源问题**。此外，在英国的每一条道路上安装低成本传感器都需要大量且不利于环境的投资。正是在这一前提下，Clytell(UK)Limited在过去12个月里一直在进行案例研究，以开发专有的**深度学习模型，预测M1和M56高速公路上的污染物浓度**。使用来自英国高速公路最先进的AQ分析仪的空气质量数据，来自Ardnance Survey API的地理信息系统位置数据，来自Open Weather API的天气数据和MIDAS的交通数据，深度学习模型能够**准确地预测(@94%)**这些高速公路上的CO、CO2、NO和NO2的污染物浓度。在可行性研究取得重大突破的基础上(即深度学习确实可以使用关键的历史数据预测空气质量)，Clytell(UK)Limited提出了一个**基于人工智能的个人和英国地区空气污染暴露(APEX)可视化平台(APEX系统)**，将把案例研究的可行性深度学习模型扩展和推广到M1和M56高速公路**，覆盖整个英国**。

Existing navigation applications (e.g. Google Maps, Apple Maps, Waze) only provide real-time traffic information despite long-term exposure to **air pollution (i.e. inorganic and organic pollutants) being the most significant environmental threat to human health (Public Health England, 2020)**. High concentrations of inorganic air pollutants (e.g. PM2.5, PM10, CO2, NO, NO2 etc.) have been associated with devastating health diseases (e.g. stroke, lung cancer, asthma) (WHO, 2016) while organic air pollutants, particularly aerosols have recently been known to be active transporters of biological agents (e.g. Coronavirus) (Wang & Du, 2020). Hence, there is a pressing and cogent need to find innovative and sustainable ways to monitor air pollutants and curb their devastating effects on the populace (DEFRA, 2020). Despite this urgency over the years, evidence suggests that **"lack of a sustainable and generalisable air quality monitoring instruments/system"** is the main reason why there is currently no publicly available system/platform that allows the UK population to navigate pollution hotspots and its devastating effects safely. Although reference air quality monitoring instruments exist (e.g. Highways England's Air Quality Analysers), they are known to be **incredibly expensive to install and maintain**; therefore, such reference instruments cannot be relatively deployed across the entire UK to capture pollutant concentration levels. Furthermore, low-cost sensors (e.g. Zephyr, AirScan) which have recently become prevalent are known to be **plagued with accuracy, reliability and power issues**. This is coupled with the fact that it would take a significant and environmentally unfriendly investment to install low-cost sensors across every road in the UK. It was on this premise that Clytell (UK) Limited have spent the last 12 months on case-study feasibility to develop proprietary **Deep Learning Models that predict pollutant concentrations along the M1 and M56 motorway**. Using Air Quality Data from Highways England's state-of-the-art AQ Analyzers, GIS Location Data from Ordnance Survey API; Weather Data from Open Weather API and Traffic Data from MIDAS, the Deep Learning models were able to **accurately predict (@94%)** pollutant concentration of CO, CO2, NO and NO2 on these motorways. On the back of the feasibility study's significant breakthrough (i.e. that Deep Learning can indeed predict Air Quality using key historical data), Clytell (UK) Limited is proposing an **AI-based Air Pollution Exposure (APEX) Visualisation Platform for Individuals and UK Regions (APEX System)** that will extend and generalise the case-study feasibility Deep Learning Models beyond M1 and M56 motorway **to encompass the whole of the UK**.