The AIQS project seeks to improve the accuracy of air quality (AQ) data from low-cost sensors and integrate this information into routing algorithms to direct pedestrians along the least polluted, greenest paths in urban areas.

To boost data reliability, the project employs two primary strategies:

• refining data accuracy through advanced analysis techniques

• evaluating hardware improvements for the sensors

Both strategies will contribute to the TRL (Technology Readiness Level) advancement of the MitH framework, specifically tailored for correcting AQ data.

Data analysis includes:

• dynamically correcting the effects of humidity on particulate matter

• identifying and fixing data anomalies

• applying various advanced methods such as Multi-Layer Perceptron (MLP) neural networks, fuzzy logic, Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Transformers, foundational models and other machine learning techniques to correct and forecast particulate matter data

Sensor hardware improvements focus on optimizing air sample efficiency and mitigating the sampling process artifacts. This involves:

• enhancing the fluid dynamics of the air collected for the sensor through simulations

• exploring energy-efficient pre-treatment options for air samples, such as dehydration modules

• testing methods for aggregates reduction using electrostatic or ultrasonic fields

Additionally, the project will enhance routing algorithms to use the improved AQ data, providing routes that minimize pollution exposure for urban commuters. By combining these efforts, AIQS aims to make low-cost AQ sensors more reliable and enable smarter, health-conscious navigation in cities, promoting a greener and healthier urban environment.

The MitH framework, currently at TRL 5, aims to reach TRL 7 by being released as an open-source solution that can be easily integrated into various low-cost sensor data publication platforms. By incorporating new anomaly detection techniques and advanced AI algorithms, MitH will significantly enhance the accuracy and reliability of AQ data correction.

Its TRL advancement will be demonstrated through extensive testing in real-world scenarios across multiple cities, including Italian and international cities, showcasing its adaptability and effectiveness in diverse urban environments.

The open-source release will facilitate widespread adoption and allow for seamless integration with existing AQ monitoring systems, further validating its readiness for operational use.

The project proposal gathers substantial interest from industrial and public stakeholders (as evidenced by the letter of interest from Wiseair and ongoing collaborations with national ARPA agencies) and includes significant female researcher participation at 43.5%.

Additionally, the project adheres to the digital requirement of ECOSISTER, with most activities and budget aligned with digitalization themes, as evidenced by the purchase of HPC servers and extensive data analysis activities.