New research, undertaken by Kingston University and network communications company Technocomm Consulting has found that AI can dramatically improve the accuracy of low cost air quality sensors.
The project, which received funding from Innovate UK, aimed to make air quality monitoring more accurate and more accessible by combining low-cost electrochemical sensors and AI technology to give accurate, real-time air quality updates from just about anywhere.
Focussing on how these sensors change over time and in different environments, the project looked at how fast they respond, how accurate they stay, how much their readings might drift, and how sensitive they are. Based on all that, the team worked on creating a machine learning algorithm that could automatically adjust to keep the data accurate over time.
The sensors were co-located alongside a reference monitor at Weybourne Atmospheric Observatory (pictured) on the North Norfolk coast.
The team took data from the both the low cost monitors and the highly accurate reference monitor over three months last summer, measuring levels of carbon monoxide, carbon dioxide and ozone at 30 minute intervals.
They fed this data into advanced AI models to create predictive algorithms that can adjust the comparative inaccuracies found in the low-cost monitor in real time.
They found that this reduced those inaccuracies by up to 46%, turning them into what the team call ‘precision tools’. The AI models can also be used on previously collected data, improving the accuracy of information collected in the past.
In their ‘success stories’ section, Innovate UK describe the project as laying: ‘a strong foundation for advancing the adoption of air-quality monitoring networks at local, national, and international levels.
‘The future of this partnership has the potential to generate intellectual property that contributes to societal well-being and economic value by revolutionising air-quality monitoring and traffic management.’
Knowledge Exchange and Research Institute Director for Cyber, Engineering and Digital Technologies at Kingston University and co-investigator Professor Jean-Christophe Nebel said: ‘We’ve discovered that portable air sensors, powered by AI, give accurate enough data to really make a difference to the public.
‘The data has the potential to inform policy decisions and enable emergency measures at local levels to directly contribute to protect the health of the public – revolutionising air quality monitoring and traffic management.
‘Our dream is to have one of these sensors on every bus or refuse collection vehicles visiting every single postcode, and for this to provide easily accessible and highly accurate air pollution data to everyone about where they live or work.’
Senior lecturer and MSc Data Science course leader and principal-investigator Dr Farzana Rahman added: ‘The innovative AI-powered sensors transform air quality monitoring and have made the data more accurate and accessible than ever. This collaboration has not only addressed a critical public health challenge but also set the stage for future advancements and impactful partnerships.’
Managing Director of Technocomm Consulting Ltd Bijan Mohandes described how the close collaboration between Kingston University and Technocomm made the project a success. ‘The regular team meetings with follow-on action items and execution were instrumental in defining the successful outcome of the project on time. The research showed that Machine Learning and AI have a role to play in modelling accurate electrotechnical sensors.’