Enhanced Wearable Single-Handed Control System Merging Inertial and Flex Sensors with Electrotactile Feedback

Abstract

Many wearable interface solutions use external computing systems—such as a computer—to infer actions from gestures, limiting feedback to a single vibrotactile or visual channel and rarely reporting on execution times or energy consumption. This paper presents the design and implementation of a wearable single-handed control system based on the Arduino Nano RP2040 Connect device, integrating an inertial measurement unit (IMU) using the Mahony orientation filter for precise attitude tracking and a set of piezo-resistive flex sensors for throttle mapping and gesture-based command execution. A TinyML neural network embedded on the Arduino device classifies hand gestures. The system also delivers multi-channel haptic feedback, creating tactile sensation via Transcutaneous Electrical Nerve Stimulation (TENS) electrodes. Operation with a real quadcopter shows concurrent four-axis continuous control, seven discrete commands, wireless transmission, and command-confirmation feedback. The embedded classifier achieves an accuracy of 0.935 and an F1-score of 0.934, with 1129 parameters and a mean inference time of 1.415 ms, matching results for the best reference classifier, while the complete control cycle remains below 8.031 ms, the sensing-to-transmission latency is below 33.1 ms, and the total glove consumption is 443.55 mW. These results establish a compact multimodal wearable platform whose control mappings can be adapted to other interactive applications such as the UAV teleoperation selected for validation in this work.