Research On Compliant Control Of Rehabilitation Robot Based On Stiffness Estimation And Fatigue Identification Of Ankle Joint

Ankle joint is very important to maintain the balance of standing and walking.The growing social aging problem has led to an increasing number of patients with ankle joint dyskinesia.Rehabilitation robots have the advantages of precise position control and low labor cost and good rehabilitation effect.It is of great significance to use rehabilitation robots for assistant training of patients to alleviate the contradiction between supply and demand of rehabilitation medical resources.Current control strategies of ankle rehabilitation focus on the recognition of patients' active motor intention,which is still insufficient in patients' active participation and the effectiveness of rehabilitation training.Therefore,it is very important to carry out the research on rehabilitation robot compliant control based on the patients' ankle joint movement information during rehabilitation training through surface electromyogram signals.Training parameters are adjusted to encourage the patients' initiative of participating in rehabilitation training,providing safe and effective rehabilitation training for patients.This thesis takes the human ankle joint as the research object.Subject-specific joint stiffness estimation and dynamic fatigue identification were deeply studied through surface electromyogram signals.Rehabilitation robot compliant control parameters were adaptively adjusted according to the movement information.The ankle joint rehabilitation robot control system was designed to realize the compliant control strategy under fatigue supervision.The main research contents are as follows:(1)It is difficult to establish a unified ankle joint stiffness estimation model because of the individual differences of patients' physiological parameters.A subjectspecific musculoskeletal geometric model in ankle plantar-dorsiflexion movement was proposed based on the study of Hill muscle tendon model.Surface electromyogram signals and joint angle signals were collected during ankle plantar-dorsiflexion movement.The denoised signals were used as the model input.The genetic algorithm was combined with forward and inverse dynamics principles to realize the subjectspecific estimation of muscle fiber length and muscle force arm and obtain the subjectspecific stiffness estimation model in ankle plantar-dorsiflexion movement.Finally,the effectiveness of the whole model was analyzed and verified by experiments.(2)To identify the fatigue state of patients' ankle joints during rehabilitation training,the muscle synergy theory was used to select the channels.The frequency domain and nonlinear features of the surface electromyogram signals of the corresponding channels were extracted to form a fusion feature.Support vector machine was used to establish the relationship model between fusion feature and fatigue state.A whale optimization algorithm based on differential evolution was proposed to determine penalty factor and kernel function parameter of support vector machine.Experiments showed that the proposed algorithm had better search ability and effectively improved the identification accuracy of fatigue state.(3)To meet the patients' rehabilitation needs in active rehabilitation training,an adaptive impedance control method based on joint stiffness estimation and dynamic fatigue feedback under fatigue supervision was proposed.In order to avoid the secondary injury of joints and muscles,the robot stiffness was adjusted to adapt to the changes of patients' ankle stiffness and the robot damping was adjusted to adapt to the changes of fatigue information,realizing rehabilitation robot compliant control.The patients' fatigue state was supervised and rehabilitation training was stopped after the detection of fatigue.On this basis,the rehabilitation robot control system was designed to visually display and record the patients' rehabilitation training data.The practicability of the system and the effectiveness of the control method were verified in the experimental environment.