What are a fog machine, strobe lights, and fake bats doing in a robotics lab? No, he's not throwing a Halloween party. It's the testbed for the next generation of rescue drones. While current models rely on cameras and laser sensors, researchers at the Worcester Polytechnic Institute They're developing mini drones that "see" with ultrasound. Like bats.
They work in total darkness, in thick smoke, in fog. They cost little, weigh less than a pound, and could save lives where traditional systems fail. Professor Nitin Sanket he just got $705 from the NSF to perfect them. Want to know more? I'm here for that.
When sight becomes a limitation
For over a decade, aerial robotics has focused on vision-based systems. Cameras, optical sensors, lidar: all work fine as long as there's light and clean air. But when an earthquake strikes at night, when a fire fills the air with smoke, when an avalanche kicks up snow dust, these systems become useless. The cameras record only black. The lasers are lost in suspended particles. The rescue efforts stop.
Sanket knows this well:
"When there's an earthquake or tsunami, the first thing to go down are the power lines. It's often nighttime. You can't wait until morning to search for survivors."
So he looked at nature (blessed be it always) biomimetics). Is there any creature in the world capable of navigating perfectly in the dark? The answer has been flitting around caves for millions of years: bats.
The mini drone that hears instead of seeing
Il PeAR Bat (from the name of the research group Perception and Autonomous Robotics) weighs less than 100 grams and measures less than 10 centimeters. It has no cameras. It has no LIDAR. It has ultrasonic sensors: the cheap ones you find in automatic faucets in public bathrooms. It emits high-frequency sound pulses and "listens" to the echoes that bounce off obstacles. Sound, unlike light, penetrates smoke, fog, and dust.
During the tests in the laboratory transformed into an artificial fog chamber, the student Colin Balfour He flew a mini drone first with the lights on, then in near-total darkness. The result? Identical. The drone identified a plexiglass wall and reversed course autonomously, without any visual aids.
La acoustic technology manages navigation and obstacle avoidance with much higher energy efficiency than optical systems.
The cameras don't disappear: they continue to be useful for locating survivors once you reach the area. But primary navigation is through sound alone.
Mini drones, the (solved) "heel" of noise
Development did not go smoothly. The noise of the propellers interfered with the ultrasound, rendering the system virtually deaf. The solution came from the metamaterialsStructures with geometries designed to modulate sound waves. A bit like sound-absorbing foam in recording studios, but applied to drones. The team has 3D printed protective shells that dramatically reduce interference.
Then there's the softwareArtificial intelligence has been trained to filter and interpret ultrasonic signals using deep learning Informed by physics. A hierarchical reinforcement learning system allows drones to navigate defined targets while dynamically avoiding obstacles. All computation occurs onboard, without the need for external infrastructure.
Beyond the bats, but not too much
Sanket is the first to admit the limitations: "Bats are amazing. We're not even close." Bats contract and compress their muscles to listen for only certain echoes, and can detect objects as thin as a human hair from several meters away.
The PeAR Bat, for now, can avoid obstacles by flying at about 2 meters per second. Slow for a real-world rescue mission, but it's just the beginning.
The project aims to develop swarms of drones that operate in environments where conventional systems are ineffective. By integrating echolocation with inertial measurement units and other sensors through the sensor fusion, these devices could dramatically improve situational awareness and navigation reliability. Future versions could also use ultrasound to detect the heartbeat of survivors, transforming drones into even more precise location tools.
Mini drones, the cost of a lifetime
A helicopter rescue mission costs up to €100.000. LIDAR is effective but energy-intensive, and in any case useless in smoke. Commercial drones with 4K cameras cost thousands of euros and become garden tools as soon as the sun goes down or the fog rolls in.
PeAR Bat is built with commercial-grade hobby components. It costs just a few hundred euros. The development team plans to move from lab testing to field deployment within three to five years.
Applications extend beyond rescue: monitoring in disaster areas, inspection of hazardous environments, and environmental protection. Sanket suggests that the principles of sound navigation could benefit sectors as diverse as self-driving cars, coral reef preservation, volcanic exploration.
The researcher is already working to increase the speed beyond 2 meters per second. At highway speeds in a forest, sounds compress: a phenomenon that must be accounted for in the models. It's not a technical detail. It's the difference between arriving on time or not.
The flight has already begun
The United States is not alone. Norway produces the Black Hornet 4, a palm-sized drone used by Western militaries. It won the award Blue UAS Refresh of the U.S. Department of Defense in 2025 for battery life and weather resistance. Harvard is working on the RoboBee, a micro drone capable of flying, landing, and even transitioning from water to air. The U.S. Air Force has confirmed the development of miniaturized drones, though without providing public updates on their progress.
These mini drones could revolutionize civilian sectors as well. Precision agriculture, infrastructure inspection, wildlife monitoring: anywhere that requires operation in challenging conditions without risking human lives.
Sanket concludes with a consideration that seems almost obvious, yet crucial: "We can't wait for the smoke to clear in emergency conditions." It's a simple statement. But it captures the reason why $705 from the National Science Foundation will be spent over the next three years on ultrasonic sensors, 3D-printed metamaterials, and neural networks trained to "sense" the world.
Millions of years of evolution have taught bats to fly in the dark. Now it's robots' turn to learn the same lesson. And when they do, perhaps they'll be able to save a few more lives.