Gas distribution mapping based on integrating system with mobile robot and sensor network

The monitoring of industrial environmental gas quality becomes an increasingly important topic due to its significant impact on the health of the producers and the safety of the production process. Traditionally, there are two main approaches to this task: first, using mobile robots equipped with sensors to patrol and monitor the environment, the prediction accuracy of this method is relatively poor; second, using a sensor network distributed in the environment for a long term monitoring, however this method lefts spatial relations and non-detection areas behind. To solve the above problems, this paper proposes a method combining the spatial map constructed by the mobile robot with fixed monitoring sensors. In this method, the concentration of gas is collected by a fixed gas sensor and predicted by a deep neural network. At the same time, a mobile robot equipped with lidar and a gas sensor is used to model the spatial construction of the environment as well as verify the prediction. By embedding the predicted concentration map with the spatial map, a gas concentration map with high accuracy and spatial information is obtained. Experiments using anhydrous ethanol odor sources in a closed and windless room showed high prediction accuracy (98%) with this method.