Manual drone control requires constantly looking at a screen while coordinating actions at the same time. Attention is divided between the aircraft and the surrounding environment. Researchers from Chonbuk University in the Republic of Korea proposed a wearable interface: commands are given by hand gestures, and the drone's position relative to the person is conveyed by vibration on the abdomen.
The work was led by Professor Ki-Ho Yu from the Department of Aerospace Engineering together with researcher Myung-Ho Shin and Dr. Giancarlo Eder Guerra Padilla. The device consists of a sensor on the back of the hand and a belt with 36 vibration motors. A computer recognizes 12 gestures as commands, and the drone's telemetry - data about its flight - is converted into vibration patterns indicating direction, relative altitude and distance.
The system was tested in several stages: gesture recognition tests, flight simulation, a study of tactile perception indoors, and two tasks with a real drone. The work was published in the journal IEEE Robotics and Automation Letters.
The gesture recognition model showed an accuracy of 97.58% and a macro F1 - an averaged quality metric across all classes - of 0.9773. In a simulated waypoint flight with 12 novices, the average completion time decreased from 263 seconds on the first attempt to 186 seconds on the fifth. Indoors, 12 participants completed 1200 attempts with the belt without a visual cue. They correctly identified direction, altitude and distance simultaneously in 82.75% of cases, and the average response time was 3.1 seconds. Direction was recognized more reliably than distance - the most difficult signal.
"The goal was to link the user's intention with clear information about the drone's position, rather than improving commands or feedback separately," says Yu. "The sensor on the hand allows commands to be given with familiar movements, while the device on the belt communicates spatial information through touch."
Waypoint tests were conducted with a real drone: one novice used the gesture controller with visual or only tactile feedback. With visual feedback, the target was reached in all nine attempts in an average of 31.4 seconds. With tactile feedback only, in eight of nine attempts in an average of 45.1 seconds. In a separate task, the participant returned the drone to the target zone after all eight exits beyond the boundaries, while vision and hearing were blocked. The results indicate a fundamental possibility rather than readiness for field conditions: only one person took part in the tests with a real drone.
"Feedback through touch is promising, but the results show where the system needs improvement, especially in conveying distance and reducing response delay," Yu noted.
Overall, the interface allows a drone to be controlled by gestures and its relative position to be felt through touch. Potential future applications are infrastructure inspection, support for search-and-rescue operations and training where it is important for the operator to monitor the surroundings. These scenarios were not tested in the study. Before assessing practical deployment, larger trials in various weather conditions are needed, as well as improvements in distance conveyance, response speed, comfort and safety.