Jeonbuk National University tests drone control using hand gestures and a vibration belt without relying on vision

A Jeonbuk National University research team developed a wearable interface that uses hand gestures to command a drone and belt vibrations to convey its direction, relative altitude and distance. The research was described in an announcement dated November 12, 2026, and the paper was published online on June 12, 2026. In real-drone point-to-point flights without vision, one novice reached the target in eight of nine attempts, with an average completion time of 45.1 seconds.

In response to the challenge of screen-based piloting diverting attention from the surroundings, this approach separates hand gestures for commands from belt vibrations for position information. The announcement also describes real-drone tests using only tactile feedback, but the test involved just one novice, and completion times were longer than with vision. Further testing is therefore needed before applying the approach to real-world operations in which users must watch their surroundings.

The key point of the research is that it assigns piloting commands and awareness of the drone’s state to separate sensory channels. Hand movements provide command input, while waist vibrations convey direction, altitude and distance. The design aims to reduce the need to keep looking at a screen, but the announcement describes an initial evaluation; it does not establish that visual information can be dispensed with. The difference in results is important when considering the system’s scope. Gesture recognition accuracy was 97.58%, while the rate of correctly identifying three spatial cues from the belt at the same time was 82.75%; the announcement said distance was the cue most likely to be confused. In real-drone point-to-point flights, the average completion time using tactile feedback alone was 45.1 seconds, longer than the 31.4 seconds with vision, and the success counts were eight versus nine. In other words, command recognition and an operator’s ability to interpret the drone’s state through touch and complete a flight need to be evaluated separately. A feature distinguishing the system from gesture control alone is its feedback path, which converts drone telemetry into vibration patterns and relays them to the operator. Indoor tests involved 12 participants and 1,200 trials in total, but both real-drone point-to-point flights and recovery from crossing a boundary were tested with just one participant. Results from simulations and indoor cue-identification tests cannot be directly applied to piloting performance outdoors in wind or around obstacles. The research team lists infrastructure inspection and search-and-rescue support as potential future applications, while noting that these uses were not tested. The next step is to broaden real-drone evaluations to include more participants and flight conditions, and assess how distance transmission accuracy, response delay, comfort while wearing the system and safety change. The current results show the potential of operating with vision blocked; they do not show that the system is ready for field deployment.

A sensor worn on the back of the hand converts 12 types of gestures into commands, while a belt equipped with 36 vibration motors conveys drone telemetry as tactile patterns. Gesture recognition accuracy was 97.58%, with a macro F1 score of 0.9773. In a waypoint task using a simulation, 12 novices reduced their average completion time from 263 seconds on the first attempt to 186 seconds on the fifth. In indoor tests conducted without visual guidance, 12 participants completed 1,200 trials in total. The rate of correctly identifying direction, altitude and distance simultaneously was 82.75%, and the average response time was 3.1 seconds. In real-drone point-to-point flights, all nine attempts with visual feedback succeeded, with an average time of 31.4 seconds; using tactile feedback alone, eight of nine attempts succeeded, with an average time of 45.1 seconds. In a separate real-drone task, the participant returned to the target area in all eight attempts after crossing a boundary, with vision and hearing blocked. The announcement said the real-drone evaluation involved one participant.