Researchers from Indian Engineering and Technology College, Vilnius Technical University named after Gediminas and Budapest University of Technologies and Economics developed an intelligent system for car emissions monitoring. It is based on the continuity principle: instead of one-off inspections at the vehicle inspection stations, the car can independently continuously transmit data about the emissions composition, and the measurements recordings are protected and cannot be changed afterwards. To achieve that, the researchers combined inexpensive IoT-sensors, machine learning algorithms and blockchain infrastructure.
The transition to the new architecture is caused by the restrictions of the existing systems. In many countries environmental monitoring comes down to a formality, despite the fact that concentrations of CO₂, nitrogen oxides, hydrocarbons and PM2.5 in big cities steadily exceed the safety thresholds. It may be seen using the example of Indian PUCC (Pollution Under Control Certificate) – the mandatory certificate, that cars must obtain every several months based on the results of the emissions inspection. In theory, this procedure stipulates for objective measurements by certified inspectors. In reality, the inspectors often use deficient or non-calibrated analyzers, enter the readings manually or perform the tests in a formalistic manner. It leads to PUCC providing for nominal concordance with the requirements, rather than real decrease of air pollution.
The researchers propose to substitute the periodical inspections with continuous digital monitoring excluding the human factor. They install a compacted block of sensors into the car, and they measure concentration of carbon monoxide, nitrogen oxides and hydrocarbons, and even minor solid particles for diesel-fueled cars. The measurements are updated every 15 minutes, and after that the owner’s smartphone does the preliminary processing, encrypts the data and transmits the into Polygon blockchain network. The recording becomes unchangeable: it is technically impossible to adjust or to delete the readings, which improves the reliability of the monitoring results.
At the same time the system changes the organization of monitoring. A network of regional inspectors responsible each for his/her own section is used instead of the centralized data storage. The inspectors change the correctness of the received data and sign digital certificates in accordance with the local regulations. This decreases the load on the infrastructure, reduces the risk of manipulation and makes the model better fit for working in major metropolitan areas. Economic incentives are stipulated for the owners of the vehicles: the compliance with the norms allows for receiving digital tokens, which provide for discounts on fuel or insurance.
The second key element is predictive analytics. The researchers trained XGBoost model using the data of the British regulator VCA, so that it can forecast excessive emissions. The algorithm takes into account the engine type, the car performance parameters and the preceding measurements dynamics. After that it defined the possible time of worsening of the measurements and of the need for maintenance. The accuracy of such forecasts reaches 98-100% transforming the system from the instrument of recording violations into the early prevention mechanism: the car owner get an opportunity to fix the deficiencies prior to exceeding the norms.
The trials using the city traffic simulator SUMO confirmed that the system was operating in the real-time-mode: the lag did not exceed 60-90 milliseconds, and the throughput capability reached 679 Mbps. From the standpoint of aggregate parameters (accuracy of forecasts, resistance to manipulations and speed of processing) the proposed architecture leaves behind the traditional and cloud solutions.
Eventually, the new solution blends seamlessly with the “smart” city infrastructure, where cars are becoming elements of the distributed digital network, and the emissions monitoring turns into automated process instead of a formalistic procedure.
