Nearly Invisible Drone Phantom Twist Developed
Northwestern University
Robotics engineers at Northwestern University have introduced the Phantom Twist drone, which becomes nearly invisible to the human eye by spinning at 15–25 Hz. The device uses the persistence of vision effect: rapid rotation turns it into a transparent blur. The design was optimized using a computational method that minimizes the LPIPS metric (Learned Perceptual Image Patch Similarity) for image similarity with the background.
Researchers from Northwestern University (Evanston, Illinois) presented the Phantom Twist drone at the 2026 Robotics Science and Systems conference in Sydney. The drone is nearly invisible to humans. The secret lies in rapid rotation: at frequencies between 15 and 25 Hz, the device exploits the persistence of vision—the human eye averages out fast motion, rendering the object as a transparent blur. The design was optimized using computational design: from 20,000 initial variants, the one minimizing the Learned Perceptual Image Patch Similarity (LPIPS) metric was selected. LPIPS measures the difference between an image of the background and the background with a simulated rotating drone superimposed. The optimized version achieved an LPIPS of 0.0104, which is about 20 times better than a human-designed version and over 10 times better than a regular quadcopter of the same size. The drone has a single motor and propeller, controlled by motor pulses at specific points during rotation, and is stabilized by rotation. Its components include a motor, propeller, batteries, controller, counterweights (which can be replaced by additional batteries or payload), carbon fiber rods, and a connector for hand launch. Currently, the Phantom Twist uses an optical tracking system and operates only in a controlled environment, but project lead Michael Rubenstein plans to adapt it for outdoor use. He also proposed mounting a camera on the rotating body for panoramic views. Potential applications include covert surveillance and wildlife observation without disturbing animals.
Source: IEEE Spectrum AI —
original
