Current state of the art including relevant previous work
Clean air is vital for people’s health and the environment. Nowadays, air pollution is proven damaging both for human health and for the environment. It is therefore important to monitor levels of air pollution, Particulate Matters (PM2.5, PM10) and NO2 to aid the management of emissions.
In vehicular cabins, there are several commercial mobile Monitoring Platforms (MPs) to monitor air quality such as SGX Sensortech [SGX19], SEAT [SEAT19], Volvo IAQS [VOLVO19]. However, in these platforms, air quality sensors are integrated into the cars air conditioning system and the platforms are not be sold separately in the market. In addition, the quality of these platforms in monitoring air quality in-cab is still questionable.
For operating vehicles, there are no sufficient existing commercial MP to monitor air quality in the cabin, especially in buses and trucks’ cabin, where the drivers spend most of their working hours, as well as large portions of their sleeping hours. Some moveable and portable instruments are used for in-cabin monitoring, but they fail to provide accurate real-time pollution information. High quality air quality measurements are done by using complex measurement techniques with sophisticated and high precision equipment with proper calibration techniques. In addition to the very high purchase and maintenance cost of these monitoring systems, they are assisted by built-in calibrator, air filtering equipment, temperature controller (cooler and heater) and relative humidity controllers, which can make the products very bulky and not easy to mount in the vehicular cabins.
It is the same situation for the use of the small size of low-cost sensor (LCS) enabling mobile applications to monitor air quality. Winsen optical sensors for PM monitoring [Báthorya17] is used for mobile monitoring that recorded and transmitted temperature (T), relative humidity (H), pressure, PM2.5, PM10 along a route in both, rural and urban area. LCS units measuring PM1, PM2.5, PM10, NO2, T and H [Firla19] are deployed as mobile sensors on ships to monitor air quality on the Baltic seas and ports.
In theory, using those small size LCSs in “mobile” air quality monitoring enables you to cover more locations, quickly and cheaper than fixed stations. Air pollutant concentrations varies in space and time providing a perfect opportunity to localise hotspots using a mobile platform to map the dynamics of air pollution: evaluating the exposure of air pollutants on individual level or improving the temporal and spatial resolution of air quality information. However, accuracy of LCSs is often questionable due to its geographical dependency (e.g., local atmospheric conditions, pollutant concentration levels) [EURO19]. To evaluate air quality sensor calibration modelling, R² is the most common metric to use. For some pollutants such as PM and NO2, supervised learning techniques (ANN- Artificial Neural Networks, RF- Random Forest, SVM- Support Vector Machine or SVR- Support Vector Regression) performed slightly better than MLRs (multi-linear regression model) looking at the coefficient of determination R² of comparison tests in field [VITO18].
As shown in Table 1, the air quality MP of AirSensEUR (v.2) developed by LiberaIntentio [Karagulian19] for detection of NO2, CO, O3, NO resulted in a mean R² value of 0.90 while the AIRQuino developed by the CNR [Cavaliere18] for detection of PM resulted in a mean R² value of 0.91. This sensor system is already operative and has undergone multiple calibrations and field tests where measurements of gaseous pollutants showing good agreement with reference measurements. Following the market analysis on [EURO19], the only sensor system satisfying the requirements of multi-pollutant, availability of raw data, transparency of all applied data treatment, availability of evaluation of the performance of sensor system in literature with high coefficient of determination (>0.85) has been found to be the AirSensEUR v.2. However, this sensor is very costly (1600EUR), not designed for in-cab and mobile solutions. HAPADS will be a solution to overcome the significant problems of the use moveable and portable air quality MPs in mobile air quality monitoring. Table 1 compares HAPADS new mobile monitoring platform with the state-of-the-art, where challenges need to be urgently addressed:
- Specific issues related to detector due to its sensitivity, detection range affected by moving velocity and wide range conditions of the outdoor weather.
- Specific issues related to data collected from mobile monitoring devices, in-motion data collection.
- Specific issues related to data storage and processing capacities on mobile platform.
- Specific issues related to data communication between vehicles and cloud/server.
- Specific issues related to calibration model to improve the accuracy for the MPs: existing MPs must be manually calibrated for a given deployment site, making them unsuitable for mobile deployment.
[AQ-SPEC15] AQ-SPEC, District, S. C. A. Q. M. & District, S. C. A. Q. M. “Air Quality Sensor Performance Evaluation Reports”, 2015.
[Báthorya17] Báthorya, C., Palotas B., “Hotspot identification with portable low-cost PM sensor”, J. Energy Water Food Nexus, Vol1, 2019.
[Cavaliere18] Cavaliere, A.; et all., “Development of Low-Cost..: Calibration and Validation of PM2.5 and PM10 Sensors”, Sensors 2018, 18.
[Corder18] Cordero et all., “Using statistical methods to carry out in field calibrations of LCS”, Sensors and Actuators B: Chemical 267, 2018.
[EURO19] European Commission, Review of sensors for air quality monitoring, JRC technical report (2019).
[Firla19] Firla S., et all., “The Influence of Marine Traffic on PM Levels…”, North and Baltic Seas, Sustainability 10, 4231, 1-19, 2018.
[Jiao17] Jiao et. all, “iScape. Summary of Air Quality sensors and recommendations for application”. iScape project D1.5, February 2017.
[Karagulian19] Karagulian, et al., “Calibration of AirSensEUR units during a field study in the Netherlands”, EC- Joint Research Centre, 2019.
[VITO18] De Vito, S. et al. “Calibrating chemical multisensory devices for real world applications: An in-depth comparison of quantitative machine learning approaches.” Sensors and Actuators B: Chemical 255, 1191–1210 (2018).