Abstract
The purpose of this study is to establish the human-exoskeleton coupling (HEC) dynamic model of the upper limb exoskeleton, overcome the difficulties of dynamic modeling caused by the differences of individual and disease conditions and the complexity of musculoskeletal system, to achieve early intervention and optimal assistance for stroke patients. This paper proposes a method of HEC dynamics modeling, and analyzes the HEC dynamics in the patient-active training (PAT) and patient-passive training (PPT) mode, and designs a step-by-step dynamic parameter identification method suitable for the PAT and PPT modes. Comparing the HEC torques obtained by the dynamic model with the real torques measured by torque sensors, the root mean square error (RMSE) can be kept within 13% in both PAT and PPT modes. A calibration experiment was intended to further verify the accuracy of dynamic parameter identification. The theoretical torque of the load calculated by the dynamic model, is compared with the difference calculated by parameter identification. The trends and peaks of the two curves are similar, and there are also errors caused by experimental measurements. Furthermore, this paper proposes a prediction model of the patient’s height and weight and HEC dynamics parameters in the PPT mode. The RMSE of the elbow and shoulder joints of the prediction model is 9.5% and 13.3%. The proposed HEC dynamic model is helpful to provide different training effects in the PAT and PPT mode and optimal training and assistance for stroke patients.
Highlights
Recent studies have shown that robots can help recover from neurological diseases such as stroke [1]–[4]
To achieve different degrees of rehabilitation training, realize the mode switching between patient-passive training (PPT) and patient-active training (PAT), and adjust the training intensity, this paper proposes the dynamic modeling and parameter identification of the whole human-exoskeleton coupling (HEC)
The accuracy of torque is verified by HEC dynamic model, and the root mean square error (RMSE) can be kept within 13% in both PPT and PAT mode
Summary
Recent studies have shown that robots can help recover from neurological diseases such as stroke [1]–[4]. The mechanism of robot rehabilitation is based on the theory of brain plasticity and human motor learning. Through repetitive motion and feedback stimulation, stroke patients can relearn feedforward and feedback motion control, promote brain reorganization and compensation, and achieve the improvement and recovery of motor function. Rehabilitation robots need to provide repetitive stimulation and enhanced feedback. As the patient’s motor function gradually recovers, it is necessary to appropriately adjust the rehabilitation robot training program. Patient-passive training is aimed at patients in the early stage of stroke. The rehabilitation robots take the patients to perform repetitive exercises to maintain joint mobility, promote blood circulation, prevent muscle atrophy, and so on. Patient-active training is the patient-guide rehabilitation training process, through robot assistance to compensate for their loss of motor ability, to complete the training task. Since active training is actively triggered and adjusted by patients, it has a stronger
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
