A big, lightweight quadcopter whose rotors turn at a constant, very low RPM, using variable blade pitch instead of motor speed for control, so the props move slowly and quietly. Machine learning designs the quietest possible blade across the full pitch range.
Rotor noise climbs steeply with tip speed, the speed at which the blade tips move. Conventional drones use small props that must spin fast to make thrust, so they’re loud. WHISPR does the opposite: big propellers turning slowly at a constant low RPM, controlled by blade pitch rather than motor speed (which responds instantly and keeps a large, high-inertia rotor precise instead of floaty).
The one honest novelty isn’t the platform. It’s ML-driven blade design optimized for low perceived noise across the entire variable-pitch envelope, validated on flying hardware, which the literature hasn’t done. Validation runs on a custom inverted-microphone test stand that measures noise vs. pitch and RPM, benchmarked against conventional drones under identical conditions.
Log
- 2026-06-01
Killed active noise cancellation
Started by trying to cancel the drone's noise in the air. A dead end: you can't reliably cancel broadband or tonal noise over an open area. Backed up to first principles instead: attack the source, not the symptom.
- 2026-07-01
Pursuing federal funding
Now pursuing federal funding through a connection with Dr. Sicheng Kevin Li at OSU's Rotorcraft Aeroacoustics Lab, who framed the ML side as a regression + classification problem on acoustic data.