A Method of Gas Source Localization from Sensor Network using Machine Learning

• Item type: Symposium(1)
• 公開日: 2020-06-17
• タイトル: A Method of Gas Source Localization from Sensor Network using Machine Learning
• タイトル / 言語: en
• タイトル: A Method of Gas Source Localization from Sensor Network using Machine Learning
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: Internet of Things
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Graduated School of Engineering, Kobe University
• 著者所属: Graduated School of Engineering, Kobe University
• 著者所属: Graduated School of Engineering, Kobe University
• 著者所属: Graduated School of Engineering, Kobe University
• 著者所属: Graduated School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduated School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduated School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduated School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduated School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduated School of Engineering, Kobe University
• 著者名: Arunothaikrit, Worachat; Ohnishi, Ayumi; Tsuchida, Shuhei; Terada, Tsutomu; Tsukamoto, Masahiko
• 著者名(英): Arunothaikrit, Worachat; Ohnishi, Ayumi; Tsuchida, Shuhei; Terada, Tsutomu; Tsukamoto, Masahiko
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Gas source localization (GSL) is one of the most important tasks to find the origin of the gas source to avoid potential danger. GSL in natural conditions is a big challenge because it has many complex conditions, especially when the wind blows in an unpredictable direction. Mobile robots use a lot of energy to work and still have short working time. That makes using an immobile sensor is better, just placing it in the right place will lead to a longer lifetime and use less energy. However, finding the location of a gas leak is difficult both on relevant and irrelevant factors. Here we show that the location and distance of the gas emission source between the gas source to the station gas sensor array in the indoor environment. We found that the machine learning algorithm is applicable to localize our experiment gas source using standard performance metrics for the regression problem in machine learning: Mean absolute error (MAE) metric. Currently, an estimated position of the source with a deviation of 3.90 cm (93.1% using R-Squared) by using Random Forests Regression (RF regression). Our results show how the stationary sensor network tends to work in finding GSLs in an indoor environment using machine learning to find the distance between the gas and the sensor in natural wind conditions. We expect our experiment to be the starting point for bringing GSL to more complex forms, for example finding distance in multiple wind direction conditions and using it on a daily basis.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Gas source localization (GSL) is one of the most important tasks to find the origin of the gas source to avoid potential danger. GSL in natural conditions is a big challenge because it has many complex conditions, especially when the wind blows in an unpredictable direction. Mobile robots use a lot of energy to work and still have short working time. That makes using an immobile sensor is better, just placing it in the right place will lead to a longer lifetime and use less energy. However, finding the location of a gas leak is difficult both on relevant and irrelevant factors. Here we show that the location and distance of the gas emission source between the gas source to the station gas sensor array in the indoor environment. We found that the machine learning algorithm is applicable to localize our experiment gas source using standard performance metrics for the regression problem in machine learning: Mean absolute error (MAE) metric. Currently, an estimated position of the source with a deviation of 3.90 cm (93.1% using R-Squared) by using Random Forests Regression (RF regression). Our results show how the stationary sensor network tends to work in finding GSLs in an indoor environment using machine learning to find the distance between the gas and the sensor in natural wind conditions. We expect our experiment to be the starting point for bringing GSL to more complex forms, for example finding distance in multiple wind direction conditions and using it on a daily basis.
• 書誌情報: マルチメディア，分散協調とモバイルシンポジウム2181論文集; 巻 2020, p. 1120-1125, 発行日 2020-06-17
• 出版者 / 言語: ja
• 出版者: 情報処理学会