Research On Control And Rehabilitation Assessment Methods Of Flexible Upper Limb Rehabilitation Robot

Stroke patients often have problems with daily activities,and they can be assisted by doctors to carry out rehabilitation training to restore their limb motor function.However,the number of physicians is limited,and patients cannot obtain continuous rehabilitation treatment.Therefore,rehabilitation robots are introduced into the field of adjuvant therapy to reduce the work intensity of rehabilitation physicians and provide patients with highintensity,repeatable and accurate exercise therapy.In rehabilitation robots with flexible ends,optical motion capture is often used to capture the motion posture of the affected limb to feedback its training status.However,this method is costly and has the problem of light spot occlusion.Based on the rope-driven flexible end upper limb rehabilitation robot,this paper uses the feedback information of inertial sensors to construct an active / passive closed-loop control system,and studies the quantitative evaluation method to assist the affected limb to realize single-joint and multi-joint active / passive training in the flexible end of the rehabilitation robot,and realize quantitative rehabilitation evaluation according to its motion information.(1)Firstly,the kinematics model of human arm and robot is constructed.The data acquisition scheme of human arm motion based on inertial sensor is adopted.According to the structure and motion characteristics of the human arm,it is analogous to a rigid connecting rod to construct a kinematic chain model with 5 degrees of freedom.By calculating the inertial sensor data in real time,the degree of freedom motion angle of shoulder joint and elbow joint during rehabilitation training is obtained.The kinematics model of rehabilitation robot and human arm is established by D-H method,and the forward kinematics and safe motion space analysis are carried out.(2)Secondly,the passive / active control system of rehabilitation robot is designed.In the passive training mode,the rehabilitation robot pulls the upper limb for training,and compares the trajectory of the degree of freedom of the shoulder joint and elbow joint with the standard trajectory.The effectiveness of the passive training action is verified by the trajectory error.In the active training mode,the motion trajectory of the elbow joint and the wrist joint in the three-dimensional space is reconstructed in real time through the degree of freedom motion angle calculated by the limb motion chain and the inertial sensor,so as to obtain the target position of each motion joint of the robot.The PID control algorithm is used to realize the dynamic tracking of the limb trajectory by the rehabilitation robot.(3)Finally,a new FMA scale score automatic prediction model is proposed.Genetic algorithm(GA)and extreme learning machine(ELM)are combined to solve the subjectivity and time-consuming problems of traditional evaluation methods.The prediction model was established by using the motion data of the shoulder joint and elbow joint related items in the UE-FMA scale and the rehabilitation physician score.The model introduces the extreme learning machine into the fitness function of the genetic algorithm,so as to extract the optimal feature subset of the patient ’s motion data in the rehabilitation assessment,and optimize the input weights and hidden layer deviations randomly generated by the extreme learning machine corresponding to the feature subset through the genetic algorithm,and realize the automation and quantification of rehabilitation assessment.The results show that the relative error of the prediction results of the model is between 0 % and 5.88 %,and the root mean square error is 0.7071,which shows good performance and effect in the quantitative evaluation of FMA scale.