AI-enabled MOS gas sensors for applications in complex environments

Traditional metal oxide semiconductor (MOS) gas sensors suffer from the common issue of poor selectivity, which restricts their widespread applications. This paper reviews the evolutionary path through which MOS gas sensors effectively enhance their selectivity and realize intelligent applications with the assistance of artificial intelligence (AI) algorithms. Firstly, the review sorts out the working mechanism and research progress of MOS gas sensors, clarifying that MOS gas sensors hold great promise for playing a crucial role in gas detection by virtue of their specific sensing mechanism, but they also face the challenge that the specific detection capability needs to be further improved in complex environments. Then, it introduces the commonly used AI algorithms in MOS gas sensor systems, covering classical machine learning algorithms such as support vector machine, random forest, principal component analysis and linear discriminant analysis, as well as neural network algorithms including back propagation neural network and convolutional neural network, which have effectively improved the sensor data processing and gas recognition capabilities. Finally, it expounds the application achievements of intelligent MOS gas sensors under the background of AI algorithms in multiple aspects, such as ensuring production safety in the field of industrial safe production, facilitating precision field operations in smart agriculture, assisting disease diagnosis and health monitoring in the field of smart medical care, and creating a comfortable environment in smart homes. The research shows that the integration of AI algorithms and MOS gas sensors has greatly expanded the application boundaries of MOS gas sensors, and meanwhile, it also points out that there are broad development prospects in algorithm optimization, hardware improvement and application expansion in the future.