Active Airflow Generation to Assist Robotic Gas Source Localization: Initial Experiments in Outdoor Environment

Abstract

Mobile robots that can search for gas sources are being developed to accomplish various tasks for protecting people’s safety, e.g., detecting gas leaks in industrial plants. If highly sensitive gas sensors are provided, robots will even be able to search for explosives set by terrorists in airports and find missing persons in disaster sites, just as trained dogs do. The use of these robots for monitoring gas emissions in landfill sites is also receiving a growing interest. In landfill sites, methane is produced by bacterial decomposition of organic waste. Since methane is flammable and has a strong greenhouse effect, methane emission monitoring is important not only for fire prevention but also for prevention of global warming. Gas molecules released in the air from their source are mainly carried by airflow, and form a plume in the downwind direction. Most gas source localization robots search for the location of a gas source by tracking a gas plume in the upwind direction. They succeeded in localizing a gas source in simplified environments, e.g., in a unidirectional airflow field in a wind tunnel. However, the direction of the airflow in the actual environments often fluctuates. Especially, the airflow direction in an open outdoor environment like a landfill site shows large fluctuations. It is difficult for the robot to track a gas plume in such an environment because the plume comes to have a complicated structure. In order to adapt to the complexity of the real environment, some robots use stochastic methods to estimate the location of a gas source. However, inaccurate modeling of the gas dispersal and/or the target environment may cause a significant drop in the performance of these robots. Instead, we propose a new robotic gas source localization method that uses actively generated airflow to regulate the airflow field. A swarm of mobile robots equipped with fans are deployed in the environment to generate a stable airflow field. A gas plume is stably formed in this modified airflow field, and therefore, a gas source localization robot can easily track the plume. In order to demonstrate the effectiveness of the proposed method, experiments were conducted in an outdoor courtyard on a moderate day. In the experiments, a robot explored on an open flat ground for a leak of flammable gas and tried to determine the direction of the leak location. The airflow velocity in this environment was 0.6 m/s on average during the experiments, and was 1.8 m/s at maximum. Four 40.8-W centrifugal fans (9BMB12PK01, SANYO DENKI) were used for active airflow generation, and were aligned on one side of a 2.8-m square experimental area. Each fan can make an airflow of 1 m/s at a point 2.5 m away from the fan. Saturated ethanol vapor was used as the detection target, and was released at a constant flow rate of 500 mL/min. The robot used in the experiments is equipped with an ultrasonic anemometer (WindSonic, Gill), and has four metal oxide gas sensors (TGS2620, Figaro) around the anemometer. Since the gas released from its source in the air is carried by the airflow, the robot can approach the gas source by moving against the wind and at the same time climbing up the gas concentration gradient. Therefore, the robot measures the airflow velocity and the gas sensor responses, and calculates the gas source direction vector by summing up the upwind vector and the gas concentration gradient vector. In the experiments, fifteen measurement points were set along two sides of the experimental area. The airflow velocity and the gas sensor responses were measured for three minutes at each measurement point. The angular difference between the actual direction of the gas source and the estimated gas source direction vector was calculated to examine the accuracy of the estimated direction. When the fans were deactivated, the estimated gas source direction vector was highly variable, although it roughly pointed to the actual location of the gas source from time to time. On the other hand, when the fans were activated, the estimated gas source direction vector pointed to the actual direction of the gas source most of the time. The results indicate that the variations of the airflow field in the outdoor environment was regulated by the airflow actively generated by the fans and that the direction of the gas source was successfully estimated in the regulated flow field. Figure 1