Evan Ackerman of IEEE Spectrum hosts Skydio CEO Adam Bry and University of Zurich robotics professor Davide Scaramuzza to discuss agile autonomous drones and their uses.
Scaramuzza says drones drew him from car research because cameras onboard could let them navigate without remote control or GPS. Their ability to cross obstacles quickly could help with search and rescue and inspections of bridges, power masts, and oil platforms. Bry describes Skydio’s goal as giving drones the sensing and software skills of an expert pilot.
Scaramuzza recounts progress from early flights at about 3 meters per second to vision-based racing at up to 72 kilometers per hour. His team’s drone matched and then beat champion Marvin in three laps using onboard vision. Tests with 20 skilled pilots measured about 220 milliseconds between a change in gaze and a throttle response, compared with roughly 20 milliseconds for the machine’s image-to-control loop.
Bry says greater agility could help in tight-gap search and rescue or difficult filming, but he sees human judgment as essential when missions depend on context. He describes current planning and control as grounded in physics-based optimization, with machine learning increasingly useful for perception and potentially more of the control stack. Scaramuzza discusses training navigation policies in simulation and says better aerodynamic physics models remain a need.
Scaramuzza explains that event cameras report pixel-level brightness changes asynchronously, which can reduce blur and latency while using little power. Bry sees them as a possible addition to standard cameras, which provide fuller scene context. On military applications, Bry says Skydio supplies dual-use drones to democratic militaries and Ukraine, while choosing not to build weapons for its aircraft.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Evan Ackerman, Mike Dooley of Labrador Systems, and Chris Jones of iRobot discuss designing useful home robots and building them with users on IEEE Spectrum’s Chatbot podcast.
IEEE Spectrum host Evan Ackerman talks with choreographer Monica Thomas and robotics researcher Amy LaViers about robot dance, their work with Boston Dynamics and the RAD Lab, and what movement can teach robotics.
