Research On Training Method Of Reconfigurable Upper Limb Rehabilitation Robot Based On Multi-sensor Fusion

With the gradual deepening of the aging process of Chinese society,the increasing demand for rehabilitation treatment every year has promoted the rapid development of the rehabilitation medical industry,and the advent of rehabilitation robots.Traditional rehabilitation robots are developed according to a specific application range,which can only perform rehabilitation training for a specific joint of the patient with movement disorder,but cannot meet the different rehabilitation needs of different patients;At the same time,for the different configurations of the reconfigurable robot,the kinematic parameters involved in the modeling process are also different.In the calibration work,all the parameters including the kinematic parameters that have little effect on the end pose are analyzed.The constructed error model will increase the amount of calculation affect the identification efficiency during the identification process;The traditional rehabilitation training methods often be used can just satisfy the patient’s mechanical trajectory training,but cannot restore the patient’s motor function and intention.Therefore,this paper proposes to assemble each joint mechanical structure of the reconfigurable upper limb rehabilitation robot simply and quickly,after that construct an exponential product kinematics model and an error model for it,then choose the sobol dimensionality reduction method to establish a dimensionality reduction optimization model,so as to complete the rapid calibration of the robot.The driven calibrated robot will establish a mirror rehabilitation training system in which the affected limb follows the movement of the healthy limb,by using electromyography sensors,force sensors and posture sensors(IMU).The main work of the thesis is as follows:(1)First of all,assemble the modules of the reconfigurable upper limb rehabilitation robot simply.This robot includes a reconfigurable shoulder joint structure,a reconfigurable elbow joint structure and a reconfigurable wrist joint structure.By combining these three mechanical structures simply and rapidly in different sequences,the rehabilitation needs of various patients can be met.(2)After combining and assembling the mechanical structure of each joint of the upper limb,the reconfigurable upper limb rehabilitation robot is modeled based on the exponential product formula and analyzed by the forward and inverse kinematics.At the same time,the sobol theory is used to calculate the global sensitivity of the robot kinematic parameters.After removing the kinematics parameters that have little effect on the end pose,the remaining parameters are used to construct the reconfigurable upper limb rehabilitation robot dimensionality reduction index product kinematics error model.(3)Aiming at the constructed dimensionality reduction exponential product kinematic error model,the LM algorithm as the identification algorithm and the laser tracker as the measurement tool are used to complete the parameter identification and error compensation of the robot,so that realize the rapid calibration of the reconfigurable upper limb rehabilitation robot.(4)According to the Fugl-Meyer(FM)motor function rating scale,14 groups of rehabilitation actions are selected.While the patient moves the healthy limbs according to the established rehabilitation actions,the movement trajectory of the healthy limbs is captured by the IMU as the desired trajectory.Then the reconfigurable upper limb rehabilitation robot after driven calibration drives the affected limb to follow the movement.(5)Finally,build a mirror rehabilitation training system that can adjust and assist dynamically,based on the multi-sensor reconfigurable upper limb rehabilitation robot.The surface electromyography signal(s EMG)and force signal of the affected limb which performing predetermined rehabilitation actions are collected simultaneously to estimate the muscle strength of the affected limb through the electromyographic sensor and the six-dimensional force sensor,thereby real-time assistance and compensation are given to the affected limb in the mirror training,so as to realize the simultaneous recovery of the patient’s motor function and intention.