Gas Source Localization Using Gas Sensor Array



Gas source localization (GSL) is a concept of searching for the source location of chemical substances drifting in the air. We are investigating GSL in outdoor environments using a 2-dimensional gas sensor array. Due to the highly turbulent nature of the outdoor wind, gas distribution in outdoor environments often has a complicated shape and show large fluctuations over time. This makes GSL in outdoor environments very challenging. To overcome this issue, we have adopted convolutional long short-term memory neural networks (CNN-LSTM) to extract spatial features from the gas distribution obtained by the sensor array and to process their time-sequential evolution. Our initial achievements obtained through the experiments conducted in a 7.5 m x 9.0 m open area with 30 gas sensors.