Northwestern's Phantom Twist spins at 1,500 RPM to evade human vision

Why it matters

Phantom Twist shows computational design moving beyond software and into the physical arrangement of machines. Its limits also separate a strong lab result from an autonomous, usable aircraft.

Michael Rubenstein, Jingxian Wang and Chen Yu demonstrated a drone on July 16th that turns its entire body into motion blur, using rapid rotation and carefully spaced components to become harder for people to see. The Northwestern University researchers presented Phantom Twist at the Robotics: Science and Systems conference in Sydney, three days before Tom's Hardware reported on the aircraft.

The drone rotates up to 25 times per second, equivalent to 1,500 RPM. Tom's Hardware incorrectly reported the speed as 25 RPM, a rate slow enough for each rotation to take more than two seconds. Northwestern's July 16th account and the researchers' paper put Phantom Twist's operating range at roughly 15 to 25 rotations per second.

That distinction explains the prototype's defining effect. At full speed, its motor, batteries, circuit board and counterweights sweep through a viewer's field of vision faster than the eye can resolve their outlines. The components visually average with the background, leaving a faint haze where a recognizable aircraft would normally hover. Phantom Twist remains visible, particularly through its wires and support rods.

Rubenstein, a Northwestern associate professor of computer science and mechanical engineering, has spent much of his career reducing robots to their essential parts. He previously helped build Kilobot, a low-cost platform used to demonstrate collective behavior across a swarm of 1,024 robots. He earned electrical engineering degrees from Purdue University and the University of Southern California, followed by a computer science PhD from USC, according to his Northwestern profile.

Wang brought a similar minimalist streak to Phantom Twist. The Peking University physics graduate previously worked with Rubenstein on PCBot, a robot built around a circuit board and a single actuator. On his research site, Wang describes his focus as full-stack robot design, minimalism, swarms and modular systems. Phantom Twist extends that work into a harder design problem: arranging the necessary parts of a flying robot around the limits of human perception.

Designing the aircraft around the observer

Phantom Twist uses one motor and one propeller. As the propeller turns in one direction, the rest of the aircraft rotates in the other. The spinning body is passively stable, while carefully timed changes in motor speed let the aircraft move horizontally. Overall thrust controls altitude.

Single-motor spinning aircraft have appeared in earlier robotics research, including maple-seed-inspired drones and Piccolissimo, a miniature vehicle developed at the University of Pennsylvania. Rubenstein's group changed the design objective. The Northwestern researchers optimized the complete physical layout for low visual perceptibility while preserving the mass distribution and aerodynamic properties required for stable flight.

That combination created a search problem with too many interacting variables for straightforward manual design. Moving a battery might reduce visual overlap from one angle while making the aircraft unstable. Shifting a counterweight could improve flight dynamics and create a darker blur against another background.

The researchers built a two-stage automated pipeline to handle those trade-offs. A computational model first generated roughly 20,000 configurations that met the requirements for stable flight. The system rendered promising designs in rotation and overlaid them on 100 real-world backgrounds. A perception model then scored how much each simulated aircraft changed the underlying image.

The researchers selected approximately 500 low-scoring candidates for further optimization. Algorithms repeatedly repositioned the motor, batteries, circuit board, propeller assembly and counterweights while enforcing the aircraft's physical constraints. The final arrangement spreads opaque components across different heights and angles, reducing the areas where those parts visually overlap during rotation.

"The design space is high dimensional," Rubenstein told IEEE Spectrum. "It's very difficult for a human to reason through all the trade-offs between the physical constraints required for stable flight and the visual appearance of the spinning drone."

Northwestern says the optimized aircraft was about 10 times less visually perceptible than a similarly sized conventional quadcopter under the researchers' metric. That result describes a model-based comparison, rather than a finding that people will miss the drone 90% of the time. The published work does not establish detection rates for unaided observers outdoors, where lighting, distance, movement and background conditions would vary.

The team fabricated and hovered multiple designs, including human-designed, randomly sampled and optimized versions. The optimized prototype received the lowest visibility score, supporting the central claim that the automated process found a physical arrangement that human designers were unlikely to reach on their own.

The lab remains part of the aircraft

Phantom Twist currently depends on an external optical tracking system and has only flown in a controlled environment. Outdoor autonomy, onboard navigation, range and endurance have not been demonstrated. The available research also leaves out the aircraft's weight, dimensions, payload capacity and cost.

Sound presents an immediate obstacle to the low-visibility thesis. The single propeller produces a noticeable whine, allowing an observer to hear a drone that may be difficult to pick out visually. Its structural rods, wiring and optical tracking tags also interrupt the blur. Rubenstein said future versions could use quieter propulsion and more transparent structural materials.

He has also proposed mounting a camera on the rotating body. A camera sweeping through every direction could provide a 360-degree view for navigation, although the researchers have not demonstrated that system. Northwestern points to wildlife monitoring, environmental surveys and infrastructure inspection as possible uses where reducing an aircraft's visual presence could limit disruption. The same property would carry clear surveillance implications if Phantom Twist developed into an autonomous platform.

Phantom Twist is an academic prototype rather than a commercial drone program. Northwestern has disclosed no company, licensing plan or production timeline. The US National Science Foundation supported the research, though Northwestern has not specified the grant amount in its announcement.

The paper was posted on May 11th, more than two months before the conference demonstration. The July 16th presentation gave the project a public stage, while the deeper result sits in the design process: Rubenstein, Wang, Yu and their collaborators gave software authority over the placement of real hardware, then fabricated the answer. The aircraft's motion blur is the visible proof that automated design can produce functional machines whose geometry follows objectives humans rarely optimize for directly.