Researchers from the University of Virginia have unveiled an innovative artificial intelligence-based system. It is able to detect toxic gases in real time, imitating the human sense of smell. The development is based on graphene with embedded nanoparticles of metal catalysts. The technology uses advanced neural networks and specialized sensors to quickly identify sources of harmful substances such as nitrogen dioxide. Its main sources in the atmosphere are car exhaust and industrial emissions. Contact with nitrogen dioxide molecules changes the conductivity of graphene, which allows to detect gas leaks with high accuracy. The system's synthetic “receptors” pick up minimal changes in pollutant concentrations and transmit the data to the sensor's computing system. Machine learning algorithms then predict the source of the leak. The neural network analyzes information from sensors placed at optimal points, which is achieved using Bayesian optimization. According to the authors of the project, this development can improve the safety of industrial facilities, urban areas and residential buildings through continuous monitoring of air quality