Robot-assisted optimized array design for accurate multi-component gas quantification

•Sensor arrays, instead of single sensors, were optimized to enhance sensing capacity.•The optoelectronic nose now enables precise quantitative analysis of gas mixtures.•An AI-robot experimental method was developed to overcome the high-dimensional challenge of array optimization. Designing sensor arrays is a common strategy for detecting mixtures. However, a sensor array designed based on human experience leads to inaccuracies due to cross-sensitivity and information overlap. This difficulty led to the realization that arrays should be optimized as a whole. The conventional array optimizing method involves selecting the best subarray from a limited sensor pool. However, this method did not consider the vast continuous multi-dimensional variable space of each sensing element’s recipe, thus only generating a simpler array with limited detection capacity. To address this problem, we developed a Robot-assisted Optimized Array Design (ROAD) method. This innovative method holistically optimizes the sensor array across a continuous and vast variable space, overcoming the bottlenecks of high dimensionality and high-quality data collection. We applied ROAD to an optoelectronic nose, tasked with detecting a mixture of CO2, NH3, and water vapor. The method explored an immense space, estimated at the order of 2^(10^7). The implementation of ROAD led to a revolutionary quantitative accuracy, achieving an average relative standard deviation of 1.90%. This advancement has propelled the application of optoelectronic nose from qualitative to quantitative. The successful ROAD of system-wide array optimization has been demonstrated in the optoelectronic nose, validating the potential of the holistically optimized sensor array for accurately quantifying each component in a gas mixture.