Strain gauges sense wind on biomimetic aerial robots
Opening up a new flight control strategy for flying robots, researchers at Institute of Science Tokyo have developed a method to detect wind direction with 99 percent accuracy using seven strain gauges on a flexible flapping wing and a convolutional neural network model.
Inspired by natural strain receptors in birds and insects, this breakthrough opens up new possibilities for improving the control and adaptability of flapping-wing aerial robots in varying wind conditions.
Flying insects and birds possess mechanical receptors on their wings that collect strain sensory data, presumably helping their flight control. These receptors possibly detect changes in wind, body movement, and environmental conditions, allowing for responsive adjustments during flight. Inspired by this natural wing with strain receptors, researchers are exploring how the wing strain sensing could extract surrounding flow information using a biomimetic flapping robot.
The study by researchers from Institute of Science Tokyo, led by Associate Professor Hiroto Tanaka, investigated the use of strain sensors on hummingbird-mimetic flexible wings to accurately detect flow directions during tethered flapping in a wind tunnel simulating hovering flight under gentle wind conditions.
“Small aerial robots cannot afford conventional flow-sensing apparatus due to severe limitations in weight and size. Hence, it would be beneficial if simple wing strain sensing could be utilized to directly recognize flow conditions without additional dedicated devices,” says Tanaka.
The researchers attached seven strain gauges, which are widely-used low-cost commercial elements, to a flexible wing structure that mimics the wings of hummingbirds. These wings were composed of tapered shafts supporting wing film similar to the structure of natural wings. The wings were attached to a flapping mechanism driven by a DC motor via a Scotch yoke mechanism and reduction gears, which generated a back-and-forth flapping motion, at a rate of 12 cycles per second. The researchers applied very weak wind of 0.8 m/s to the mechanism in a wind tunnel. The wing strain was measured during flapping under seven different wind directions (0°, 15°, 30°, 45°, 60°, 75°, and 90°) and one no-wind condition. A convolutional neural network (CNN) model was used for machine learning of the strain data to classify these wind conditions.
A high classification accuracy of 99.5 percent was achieved using the strain data with the length of a flapping cycle. Even with shorter data length of 0.2 flapping cycles, the classification accuracy remained high at 85.2 percent. Using only one of the strain gauges, the classification accuracy was also high, ranging from 95.2 percent to 98.8 percent with a data length of a flapping cycle, while the classification accuracy drastically dropped to 65.6 percent or less with the short 0.2 cycles data. These results suggest that wing strain sensing at multiple locations can enable wind direction recognition with high accuracy in as little as 0.2 flapping cycles.
By removing the inner wing shafts, the classification accuracy decreased. The degree of decrease was 4.4% with 0.2 cycles data and 0.5 percent with 1 cycle data when all strain gauges were used, respectively. Additionally, when using only one strain gauge, the decrease averaged 7.2 percent for 1 cycle data and 6 percent for 0.2 cycles data. These results suggest that the biomimetic wing shaft structures enhance the wind sensing capabilities of the wings of aerial robots.
“This study contributes to the growing understanding that hovering birds and insects may sensitively perceive wind through strain sensing of their flapping wings, which would be beneficial for responsive flight control. A similar system can be realized in biomimetic flapping-wing aerial robots using simple strain gauges,” concludes Tanaka.
Image: Biomimetic flapping-wing aerial robots in hover can accurately detect winnd direction using wing strain sensing. Credit Institute of Science Tokyo.
DOI: https://dx.doi.org/10.1002/aisy.202400473
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
