| Stroke is a common acute cerebrovascular disease.After the stroke,most patients will have upper limb dysfunction.At present,the main way of functional rehabilitation of patients is rehabilitation training.Studies have shown that,compared with traditional rehabilitation training methods,robot-assisted training can provide a multi-modal training program,effectively improving the rehabilitation training effect of stroke patients.Since the upper limb rehabilitation training of the human body is more complicated than the lower limb training,and the movement function recovery of the limbs is slower,the research and development of the upper limb rehabilitation robot have significant research significance and clinical value.At present,in the structural design of upper limb rehabilitation robots,exoskeleton-type upper limb rehabilitation robots are widely used because they can better correspond to the skeletal structure of the limbs and can directly control specific joints to complete specific rehabilitation training actions;in the control algorithms of rehabilitation robots,the human joint torque estimation algorithm based on surface electromyography(sEMG)can better realize the human-computer interaction function,so it has been extensively studied.However,there are still some shortcomings that need to be strengthened: 1.The structural design of the pronation-supination degree of freedom of the forearm of the rehabilitation exoskeleton is mostly driven by the direct-drive method of the joint of the motor attached to the exoskeleton,which will increase the exoskeleton inertial parameters affect the overall control accuracy.2.The elbow joint moment estimation algorithm is data-driven.The existing research only collects data under a certain forearm posture,and its generalization ability is insufficient when applied under different forearm postures.Aiming at the above two shortcomings,this paper proposed a joint torque estimation method based on sEMG signals and designed an elbow joint rehabilitation exoskeleton system and its control strategy.The main research work and contributions of this paper are as follows:First of all,according to the shortcomings of the joint direct-drive method,a differential form of elbow joint rehabilitation exoskeleton based on cable-driven is designed.The motor and other components are arranged on the outside of the human body based on cable-driven.The two degrees of freedom movement of elbow flexion-extension and forearm pronationsupination are realized in a differential manner.At the same time,the exoskeleton sensing system,data acquisition system,and control system are constructed.Secondly,aiming at the shortcoming of the weak generalization ability of conventional algorithms,a joint torque estimation method based on the sEMG signal is proposed.Designed a joint torque regression algorithm using BP neural network,and designed a data collection experiment by analyzing the influence of forearm posture on elbow flexor muscle force.Therefore,the model has strong generalization when applied in different forearm postures.Finally,based on the elbow joint rehabilitation exoskeleton platform and joint torque estimation method,a passive control strategy based on position control and a power assist control strategy based on torque control is proposed.The experimental results showed that the exoskeleton has better dynamic response characteristics,trajectory tracking capabilities,and plays a better role in assisting and delaying fatigue.The exoskeleton has potential application value in community rehabilitation and home-based rehabilitation. |
