AI Robot Tracks Divers’ Breathing Through Underwater Bubbles

Respiration rate can signal stress, hyperventilation or exhaustion in scuba divers, but monitoring it underwater is difficult. Thick wetsuits and drysuits can prevent wearable medical sensors from maintaining reliable contact, while wireless transmission is constrained below the surface. Researchers at the University of Minnesota Twin Cities developed a camera-based AI system that estimates breathing from exhaled bubbles, allowing an autonomous underwater vehicle to act as a safety partner without requiring divers to carry additional sensors.

The university announced the work on July 24, 2026, after researchers trained the system using thousands of manually categorized images and synchronized audio of regulator exhalations. Data were gathered in Lake Superior, Minnesota’s Square Lake and the Caribbean Sea off Barbados. The HREyes communication system classifies breathing as below normal at fewer than 14 breaths per minute, normal at 14 to 20, or above normal at more than 20, and can alert the diver in real time. The study was published in The International Journal of Robotics Research.