Artificial olfactory systems, inspired by the human sensory system, can detect and discriminate a wide variety of gas molecules, which are in resonating demand in real life applications. The two core components of artificial olfactory systems are the chemical sensors as the sensing units to perceive chemical compounds and machine learning algorithms to recognize the gas response patterns for olfactory perception. Among the sensing techniques, the nanostructured metal oxide gas sensor array outstands itself for simple structure, miniaturization, sensitivity, and robustness. However, cross-sensitivity to a broad range of gas molecules hinders its development as a platform for general artificial olfactory systems. In this thesis, integrated nanostructured-metal oxide gas sensor arrays, com...[ Read more ]

Artificial olfactory systems, inspired by the human sensory system, can detect and discriminate a wide variety of gas molecules, which are in resonating demand in real life applications. The two core components of artificial olfactory systems are the chemical sensors as the sensing units to perceive chemical compounds and machine learning algorithms to recognize the gas response patterns for olfactory perception. Among the sensing techniques, the nanostructured metal oxide gas sensor array outstands itself for simple structure, miniaturization, sensitivity, and robustness. However, cross-sensitivity to a broad range of gas molecules hinders its development as a platform for general artificial olfactory systems. In this thesis, integrated nanostructured-metal oxide gas sensor arrays, combined with machine learning algorithms, are explored for gas discrimination.

Throughout the whole work, a 3D nanotube array based on a unique nanohoneycomb structure is constructed with conformal and dense sensing film for superior sensing response toward ppb and sub-ppb gases. Next, we addressed the crosssensitivity issue from several aspects, including materials synthesis, heating modulation, sensor array construction, and algorithm design.

Starting from the rationally designed sensing materials, the platinum nanoclusters decorated SnO₂sensors achieved highly sensitive and selective NO₂ sensing, which is validated by density functional theory and experimental tests. Following the single sensor-based strategy, a micro-heater integrated sensor working on pulse heating mode achieved multi-gases discrimination with rich features during these heating and cooling processes. Multiple gases can be well-recognized with the aid of the conductance-slope map. The gas sensor array is another possible method for multi-gases discrimination. We combined principal component analysis and support vector machine to demonstrate indoor air quality monitoring with self-powered sensor array networks. Moreover, we proposed a monolithically integrated 3D sensor array chip with up to 10,000 individually addressable sensors in the millimetre chip region. Augmented with neural network algorithms, the biomimetic olfactory system empowered by sensor array chips fulfilled excellent distinguishability for more than 4000 samples of multiple gas species and concentration prediction of gas mixtures with the ultralow absolute relative error of less than 2.7%. We also demonstrated the capability of the biomimetic olfactory system to discriminate twenty-four odours. Finally, we have demonstrated the power of the fusion of vision and olfaction senses on a quadrupedal mobile robot for a mini reconnaissance mission, which uncovered the alluring potential of our biomimetic olfactory system for many future vital applications.