Echolocation for Drones

Researchers at Worcester Polytechnic Institute (WPI) have developed a low-power ultrasonic sensing system that uses a dual sonar array and neural network signal analysis to guide drones in obscured environments where radar, lidar, and GPS are impractical. The open-source prototype successfully navigated simulated forests, smoke, and snow, and was tested in actual wooded areas.

Bats are remarkable creatures, able to fly at night or inside the confines of caves without light to guide their way. A team of researchers at Worcester Polytechnic Institute WPI has determined how to use low power ultrasonic sensors to guide drones in obscured environments https://pear.wpi.edu/research/saranga.html . While radar, lidar, and GPS are all great for navigation and sensing, they can run into issues when light is obscured or can take too much power to be practical for the limited battery life of a drone. The researchers found that a dual sonar array could be used to implement a much lower power sensing system for a drone that performs well in environments that would stymie a computer vision system. A shield placed behind the array cuts down on the sound of the propellers that would otherwise drown out the signal, and further signal analysis via a neural net separates the echoes of objects in front of the drone from the background. The prototype could navigate in various simulated environments like forests, smoke, and snow. It looks like it even got a chance to go for a flight in the actual woods. All the code and hardware designs are Open Source, https://github.com/pearwpi/Saranga so have at it We’ve covered mosquito-inspired drone sensors https://hackaday.com/2020/06/17/obstacle-avoidance-for-drones-learned-from-mosquitoes/ before, and if you want to get into echolocation yourself, apparently humans can learn to do it too https://hackaday.com/2024/11/26/humans-can-learn-echolocation-too/ .