A palm-sized drone built at Worcester Polytechnic Institute is navigating fog, smoke and pitch-black rooms using the same trick bats have used for millions of years: short ultrasound chirps and a careful listen for the echoes that come back. The work, published in Science Robotics, points to a lighter, cheaper alternative to the lidar and radar systems that weigh down most small autonomous aerial robots.
The team is led by Nitin J. Sanket, assistant professor in WPI’s Department of Robotics Engineering. His pitch is simple. “Bats that weigh less than two paper clips can accurately navigate in dark, damp, and dusty caves by sending out short chirps and listening to the weak echoes with a limited number of neurons,” Sanket said. If a bat can do that on almost no compute, a drone should be able to as well.
An X-shaped quadrotor about six inches wide, kitted out with just two ultrasound sensors and an acoustic shield to muffle the noise from its own propellers. A deep learning model does the rest, trained to pick useful echo patterns out of the noise the way a bat’s brain does. The whole robot weighs about a pound and can fly for roughly five minutes on one battery.
Lidar, radar and camera-based systems all come with costs, both literal and physical. Lidar is heavy and power-hungry. Cameras struggle in darkness and bad weather. Propeller noise tangles up sound-based perception. For search-and-rescue work inside collapsed buildings or smoke-filled corridors, every gram of sensor hardware is a gram of battery the drone does not carry, and every second of flight time is a second closer to reaching someone.
Across 180 test runs, the drone cleared its courses successfully between 72% and 100% of the time. Tests included wooded outdoor flights and indoor obstacle courses with transparent plastic, metal poles and artificial fog or snow. Thin targets gave the system the most trouble; metal poles and slender tree branches reflected too little sound to register reliably.
Sanket’s team, supported by a National Science Foundation grant, now wants to shrink the platform further and push flight times higher. The argument for doing so is blunt.
“In a real search-and-rescue mission, a few more seconds of flight time could mean the difference between life and death for a survivor,” Sanket said.
Article & Image Source: Worcester Polytechnic Institute
