Human Intention Detection for Upper-Limb Assistive Exoskeletons Using a Motor's Built-in Current Sensor
Conference
(2025). Human Intention Detection for Upper-Limb Assistive Exoskeletons Using a Motor's Built-in Current Sensor
. -111. 10.1109/AIAC68175.2025.11332347
(2025). Human Intention Detection for Upper-Limb Assistive Exoskeletons Using a Motor's Built-in Current Sensor
. -111. 10.1109/AIAC68175.2025.11332347
Exoskeletons can effectively reduce physical burden, yet their active control critically depends on detection of human motion intention. Existing approaches based on human-machine interfaces, bio-signals, or force/inertial sensors face limitations of high cost, susceptibility to noise, and wearing inconvenience. To address these challenges, this work proposes an intention detection framework that relies solely on the motor's built-in current sensor. Two representative algorithms-amplitude-based threshold detection (ATD) and pattern-based template matching (PTM)-are designed and compared, with particular focus on distinguishing real intention from human-induced disturbances. Experiments on a single-degree-of-freedom elbow exoskeleton across varying joint angles and loads demonstrate that ATD achieves higher precision, recall, and F1-score, showing strong robustness to disturbances. In contrast, PTM offers better adaptability and comfort but suffers from higher false detections under extreme conditions. These findings validate the feasibility of current-sensor-based intention detection and highlight future directions such as adaptive thresholds and multimodal fusion for enhanced robustness.
