Multi-degree Of Freedom Hand Rehabilitation Robot Based On Intelligent Human-machine Interaction And Control

Stroke has the characteristics of high mortality and high disability rate,which has the greatest impact on human health.One of its main sequelae is the significant reduction of unilateral hand function,which seriously affects the independent living ability of patients.The existing hand rehabilitation mainly depends on the unarmed auxiliary training of rehabilitation therapists,which is not only time-consuming and laborious,but also affected by the therapist’s experience and the training quality is extremely unstable.The current number of rehabilitation therapists cannot match the increasing number of patients,many patients need to wait a long time to get extremely limited rehabilitation training opportunities.Therefore,the rehabilitation training effect is poor.Robot assist rehabilitation training is considered to be an effective way that can effectively replace manpower and promote post-stroke rehabilitation.At present,the rehabilitation robot has good effect on large joints of upper and lower limbs training,and has gradually replaced the unarmed training of therapists.However,the therapeutic effect of hand rehabilitation robot is not ideal.On the one hand,the human hand has a highly dexterous mechanical structure and highly fine motion function,so copying the reciprocating training mode of the large joints of the upper and lower limbs cannot effectively repair the dexterous operation function of the hand;On the other hand,the existing hand rehabilitation robots have some problems,such as less active degrees of freedom,low human-machine interaction performance,unable to fine regulate the fingertip force and position,which are difficult to meet the rehabilitation training needs of hand fine motor function.This paper focuses on the mechanical structure,control system and human-machine interaction system of hand rehabilitation robot.A multi-degree of freedom full drive hand rehabilitation robot is designed and the dexterous and fine auxiliary motion training of each finger are realized.For the practical application of rehabilitation robot,the human-machine interaction system based on surface electromyography(sEMG)is studied;Motor skills are learned from healthy people and are transfer to rehabilitation robot.Assist-as-needed controller is designed to provide accurate finger strength training for patients in combination with human-machine interaction;A controller based on radial basis function neural network(RBFNN)is designed for finger precise position rehabilitation training.The specific work is as follows:(1)Mechanism design of multi-degree of freedom full drive hand rehabilitation robot.In this paper,a multi-degree of freedom full drive hand rehabilitation robot is designed.The micro cantilever structure is used to realize the traction of each fingertip.Each joint of the micro cantilever can be adjusted to adapt to different hand sizes,making it more convenient for patients to wear.Each micro cantilever has 4 active degrees of freedom and can drag each finger to achieve bending/extension and adduction/abduction.In order to increase the perception performance of the robot,two force sensors are placed at the fingertip of the exoskeleton,which can detect the precise force on the abdomen and back of the fingertip respectively,increase the accuracy of human-machine interaction force control,and fully ensure the safety of users.The hand rehabilitation robot is designed by SolidWorks software and realized by 3D printing;The micro motor placed at the micro cantilever joint can realize 20 active degrees of freedom,and the force sensor is placed at the fingertip to realize the fine regulation of the position and force of the five fingertips.Through the forward and inverse kinematics analysis and dynamic analysis,it is found that the hand rehabilitation robot has high flexibility and good stability.(2)Research on sEMG signal decoding and motion pattern classification for human-machine interactive control of rehabilitation robot.The motion intention extraction based on sEMG can objectively and accurately reflect the user’s intention.The motion intention extracted from sEMG can control the hand rehabilitation robot in real time.In this study,sEMG signals were extracted from 18 young healthy subjects.Six muscles of forearm and hand are collected,which are palmaris longus,radial wrist flexor,total extensor digitorum,superficial flexor digitorum,abductor pollicis brevis and first dorsal interosseous muscle.A total of 22 characteristic parameters were extracted,including 15 time domain characteristic parameters,3 frequency domain characteristic parameters and 4 time-frequency domain characteristic parameters.The multicollinearity characteristic parameters were removed by correlation analysis.Then the random forest and principal component analysis were used for feature selection.Finally,6 characteristic parameters were obtained,namely root mean square value,average absolute value,slope change rate,Willison amplitude Mean frequency and 6th order autoregressive model.Three classifiers,multi-channel convolutional neural network,multi-layer perceptron and support vector machine are used for classification.The results show that the accuracy of the three classifiers for 13 commonly used rehabilitation training actions,such as single finger bending,thumb and other fingertips pinching,five finger grasping and so on,is more than 95%.(3)Research on finger strength training control system based on assist-as-needed controller.When the patient recovers certain motor function,it is necessary to carry out finger strength training according to the patient’s rehabilitation to improve the completion rate and stability of exercise.In order to provide targeted rehabilitation training according to the rehabilitation status of patients,transfer learning is used to transfer human kinematics skills to hand rehabilitation robot.Index finger bending,thumb palming and thumb bending are selected as motion trajectories,and the kinematic data of healthy subjects are collected by Motive.The training time warping algorithm is used to regularize the data of single person.Finally,the Gaussian mixture model and Gaussian mixture regression algorithm is used to fit the reference trajectory of the rehabilitation robot from the data of multiple subjects.The assist-as-needed controller can judge the active and passive motion modes by setting the fingertip force threshold and realize the adaptive switching between the modes.In the passive training mode,the rehabilitation robot drives the hand for rehabilitation training,and in the active training mode,the rehabilitation robot will carry out auxiliary motion training according to the motion intention of the hand.In the active training mode,the speed of the rehabilitation robot is determined by the human fingertip output force.When it is difficult for the user to complete the action independently,it will change from the active training mode to the passive training mode and retain the speed of the previous time to ensure the smoothness of the motion.Taking subject H1 as an example,the fingertip motion trajectory obtained by transfer learning is very similar to its actual motion trajectory,which shows that the control system can well transmit the kinematics information of the human hand to the hand rehabilitation robot,and has the ability to provide targeted rehabilitation training according to the rehabilitation status of the patient.(4)Research on fingertip fine position control method based on RBFNN controller.In this study,the mirror control strategy is adopted,the kinematics information of the healthy hand is collected through LEAP Motion,and the joint angle of the hand rehabilitation robot is calculated through inverse kinematics.The rehabilitation robot adopts RBFNN controller in fine motor rehabilitation.The simulation results show that the controller can track the desired trajectory very well,and the tracking error is less than 0.01 rad.The established RBFNN controller has good performance and can realize the fine position control of the rehabilitation robot,which is of great significance to the recovery of patients’ fine motor function in the middle and later stage of rehabilitation training.(5)Design,implementation and function test of multi-degree of freedom full drive hand rehabilitation robot.Based on the above work,a multi-degree of freedom full drive hand rehabilitation robot prototype is developed and completed.In order to verify the structural design of the rehabilitation robot,the grasping test is carried out.The finger strength training control system based on assist-as-needed controller and the fingertip fine position control system based on RBFNN controller are tested on healthy people.The results show that the multi-degree of freedom full drive hand rehabilitation robot can grasp the objects with 8 different shapes stably;The assist-as-needed controller can smoothly switch the auxiliary mode according to the needs.In the active training mode,the rehabilitation robot can adjust the motion speed according to the interaction force of the fingertip;In the actual control,the 15 joint angles controlled by RBFNN controller can well track the desired joint angle.Combined with the traditional controller,20 joints coordinate movement to realize the fine position control of fingertip.By designing and manufacturing a multi-degree of freedom full drive hand rehabilitation robot,this study designs and develops a human-machine interaction control control system based on sEMG,a finger strength training control system based on assist-as-needed controller and a fingertip fine position control system based on RBFNN controller.Unlike the predecessors who can only provide assistance for patients in the early or middle stage of rehabilitation,the multi-degree of freedom hand rehabilitation robot system based on human-machine intelligent interaction and control in this study can provide the necessary assistance to patients in all rehabilitation stages.Hand posture training based on sEMG human-machine intelligent interaction is provided for patients in the early stage of rehabilitation training,and finger strength training and fingertip fine movement training are provided for patients in the middle and late stage of rehabilitation.It lays the foundation of structural design and fine motion control for the future rehabilitation of hand function.The work of this study has greatly improved the fine training level of hand function rehabilitation robot,which is of great significance for the fine motor function recovery of hand after stroke and a variety of neuromuscular diseases or injuries,it improves patients’ hand function and the ability to return to society.