Using Air Curtains in Gas Sensing System to Enhance Its Directivity

Over the past decades, there has been growing interest in developing mobile robots that can localize gas sources. Potential applications of such robots include searches for gas leaks and environmental pollutants. The searches for hazardous chemical sources can be too risky for human workers, and therefore, are better suited for robots. However, to find a gas source is not a straightforward task for robots. Molecular diffusion is generally an extremely slow process. Therefore, the gas molecules released from their source are carried by airflow and trail in the downwind direction. The resultant aerial trail of the released gas is called a plume. Since the plume has a shape elongated in the downwind direction, the gas concentration gradient along the airflow direction is extremely small. Moreover, the airflow in environments of practical interest is almost always turbulent. Since the turbulence of the airflow makes the plume meanders randomly, the measurement of the small gas concentration gradient becomes even harder. The airflow direction can be used as a reliable cue to estimate the direction of the gas source. However, a bulky ultrasonic anemometer is required to accurately measure the direction of the weak airflow in indoor environments.

Here we propose to use actively generated air curtains to enhance the directivity of the gas sensors on a mobile robot. We have developed an active stereo nose, which consists of two gas sensors and a single air nozzle. Air is sucked using pumps into laterally aligned two intake pipes in which the gas sensors are placed. The exhaust of the pumps is ejected from the nozzle, which is placed between the two intake pipes. An air curtain generated from the nozzle has been found effective in enhancing the difference between the two gas sensor responses. The direction of a gas source can be determined simply by comparing the left and right gas sensor responses. Here we show the extension of this method to two-dimensional directional determination. The proposed system has four gas sensors, and therefore, is termed active quad nose (AQNose). A number of small nozzles are aligned to form air curtains between the front, left, back, and right gas intake pipes.

In the experiments, saturated ethanol vapor was released from a gas source at a constant rate of 100 mL/min. Two arrays of DC fans were placed in the room to generate a circulating airflow field mimicking the natural convection. The velocity of the airflow near the floor was 10 cm/s, and was almost uniform over the entire floor. A metal oxide gas sensor (TGS2620, Figaro Engineering) was placed in each of the four intake pipes to detect the ethanol gas. Air was sucked into each intake pipe at 3.0–4.4 L/min. The sensor response was defined as the ratio of the sensor resistance in gas to that in air. Since the sensor resistance decreases with the gas concentration, the sensor response value also decreases with the gas concentration. AQNose was placed at various locations from 50 cm upwind of the gas source to 200 cm downwind. At all locations tested, the gas sensor facing to the gas source showed the largest response change, and the direction of the gas source was successfully determined by comparing the response values of the four sensors. The results show the effectiveness of the active air curtain generation.