Real-time Assistance Control of Hip Exoskeleton Based on Motion Prediction

To satisfy the daily walking needs of the individuals with weak muscle strength in lower limbs, a wearable hip exoskeleton is designed to realize accurate perception of the human daily walking intentions and real-time assistance of hip flexion/extension. A motion prediction model based on hybrid oscillators is proposed to realize accurate assistance timing in the frequent stop-and-go condition and no-delay assistance for regular walking gait. Based on the motion prediction model, human walking gait prediction and gait cycle segmentation are realized by converting gait data from the time domain to the phase domain. The phase angle is used to calculate the assistance torque online, and thus the assistance control of hip exoskeleton is implemented. To eliminate the jitter caused by large assistance torque, the rigid connection between the inertial measurement unit and the hip exoskeleton thigh link is decoupled. Treadmill experiments and free walking experiments are conducted to verify the adaptability of the hip exoskeleton to sudden or uniform speed changes, and the effectiveness in stop-and-go switching or free walking conditions respectively. Meanwhile, the human-machine interaction force is detected and compared to verify that the adaptive oscillator mode with prediction ability can provide more compliant walking assistance. The experimental results show that the real-time assistance control based on the hybrid oscillators model can generate the assistance for non-periodic gaits in stop-and-go condition and no-delay assistance for quasi-periodic gait in regular walking condition.