AI-Powered Sonar Enables Small Drones to Navigate Like Bats

Researchers have developed a novel ultrasound-based perception system allowing miniature drones to navigate in low-visibility environments, drawing inspiration from the echolocation abilities of bats. Current drone navigation often relies on cameras or lidar, which are unreliable in conditions like smoke, fog, or darkness. This new system overcomes a significant challenge: the noise generated by drone propellers interfering with the ultrasound sensors. The team employed a physical acoustic shield, modeled after bat ear cartilage, to reduce propeller noise and a neural network called Saranga to recover weak echo signals. This combination enables the drones to estimate obstacle locations in 3D with a low-power sensing capability. The technology promises to revolutionize search and rescue operations, combating poaching, and exploring hazardous environments where traditional sensors fail. The new system boasts significant improvements over existing solutions, potentially reducing power consumption by 1,000 times, weight by 10 times, and cost by 100 times. Future research will focus on improving speed, range, and system size, with the ultimate goal of deploying swarms of these robots for complex missions.