Thanks to echolocation, this small drone can navigate even when it is dark or very foggy.
Drones usually ‘see’ with cameras and LIDAR sensors. In general, they can navigate well with this, but there are situations in which this does not work well. In the dark, for example, or in dense fog. There are solutions for this, for example by using radar sensors. But these types of systems are often expensive, heavy and require a lot of power – which is inconvenient for small, light drones without large batteries.
Many bat species use echolocation to navigate in the dark. They produce high-pitched sounds and listen to the echoes. From this they can determine where obstacles are located. Perhaps echolocation is also suitable for light drones, researchers from the Worcester Polytechnic Institute (USA) thought. The Thai pig-nosed bat (also known as the bumblebee bat), weighing only 2 grams, can also fly through dark caves without colliding with anything.
Also read:
Drone makes noise
They got to work with the PeARBat160, an autonomous quadrotor that is about the size of your palm and weighs 460 grams. But equipping a drone with echolocation turned out to be not that easy. Such a flying machine itself produces a lot of noise, which drowns out the relatively weak echoes. The researchers came up with two solutions for this.
First they built a sound shield on the front of the drone. This forms a physical barrier between the noisy rotor blades and the echo sensor. As a result, noise pollution was a lot less, but not completely disappeared.
The researchers therefore also developed a neural network (a form of artificial intelligence) called Saranga. Neural networks are good at recognizing patterns. The researchers taught Saranga to recognize weak echoes in noisy sound recordings. They then integrated the neural network into the quadrotor. The navigation system used only 1.2 milliwatts (radar sensors often use tens of watts).
Rescue operations
Then it was time to test the drone. During indoor and outdoor experiments, the researchers showed that the quadrotor could avoid both trees and thin wires. The drone also continued to navigate well in the dark, thick fog or when it was snowing. The small flying machine even recognized transparent ‘walls’ made of thin plastic sheets – something that both cameras and radar sensors have difficulty with.
The researchers admit that radar works better in many cases. It ‘sees’ objects from further away, allowing the drone to respond sooner and therefore fly faster. But if a drone needs to be cheap and small, echolocation (sonar) is a good alternative.
According to the researchers, robots with echolocation can also provide a solution if there is poor visibility during search and rescue operations, for example during forest fires, in collapsed buildings after earthquakes or in dark caves.
In video below you can see the drone in action:
Sources: Science Robotics, AAAS via EurekAlert!
