Locomotion Joint Angle and Moment Estimation with Soft Wearable Sensors for Personalized Exoskeleton Control

In recent years, significant advancements in flexible sensing and machine learning have led researchers to explore soft sensors as alternatives to traditional inertial measurement units (IMUs) for detecting human posture. However, much of this research has concentrated on calibrating soft sensors, with limited progress in applying these calibrated sensors to practical applications. Addressing the aforementioned issues, this study not only calibrates a pair of homemade soft-sensor-based smart knee pads using machine learning techniques, but also uses it to estimate hip joint moments during walking, which can help guide the assist torque profile for flexible exoskeleton. To estimate knee joint angles and hip joint moments with high precision using the smart knee pads, we collected and calculated gait data from 8 participants under 4 trials using the GRAIL System. The data includes bilateral knee joint angles and bilateral hip joint torques. LSTM and CNN models were specifically chosen to estimate knee joint angles and hip joint moments, respectively, with a Mean Absolute Error (MAE) of 4.43° for knee joint angles, and an MAE of 0.12 Nm/kg for hip joint moments. Subsequently, a flexible exoskeleton was developed to provide assistance based on real-time estimation of hip joint moments, enabling personalized control of the exoskeleton. Exoskeleton assisted walking trials with 5 volunteers at various speeds verified the performance of soft sensors and flexible exoskeleton. The results, demonstrated through muscle synergy and activation patterns derived from multi-channel surface electromyography (sEMG) signals, along with participant feedback, highlighted significant improvements in mobility and comfort. By using estimated joint moments as a key factor in exoskeleton assistance decisions, this research not only enhances the comfort of the exoskeleton but also represents a significant advancement towards personalized control of assistive exoskeleton.

Recommended citation: 123
Download Paper