The Control System Design Of Lower Limb Exoskeleton Based On Neural Network

Lower limb exoskeleton is a kind of wearable boosting device which imitating the physiology of human,could help the wearer improve their walking stamina,load-bearing capacity and some other physical abilities.Rehabilitation Lower limb exoskeleton is a hot area of research for the past few years at home and abroad.Meanwhile,rehabilitation Lower limb exoskeleton combine the advantages of human and machine together,and meet the expected movement through the collaboration of human and machine,then,help the patients with lower limb injuries and hemiplegia rebuild their ability to walk.Control system is one of the most important module of lower limb exoskeleton robot,the performance of it will have effects on the rehabilitation train result directly.Artificial neural network has the capacity to self-learning and self-adaption,applying neural network to control system could improve the response speed,accuracy,following features,etc.of the whole control system.Then,improve the control system's intelligence of lower limb exoskeleton and the validity of rehabilitation training.In this master thesis,the 3D Motion Capture system was used to gather the gait data of lower limb,which was used as the data support of control system.Based on the TUST 3 mechanical structure of dynamic lower limb exoskeleton which has already finished,this thesis analyzed the kinematics of the lower limb exoskeleton,built the control system and fulfill the design of lower limb exoskeleton control system in term of human lower limbs'movement characteristics in different gradients(0,1:10,1:12).The main content that has been finished in this master thesis are showed as follows:Applying 3D Motion Capture system called VICON to collect the lower limb gait data of normal subjects during walking on different gradients of slope,and after analysis process,took the data that exported from VICON system as original data set.According to the analysis of human gait in different gradients of slope and the analysis of kinematics and dynamics of lower limb exoskeleton,the neural network control system was built,including the degree of freedom of lower limb exoskeleton,the inputs and outputs of the neural network.Applying the traditional control mode to control each single joint of lower limb exoskeleton based on human lower limb biomechanical characteristic,and simulated in MATLAB.Then,observed the control effect of traditional control mode and summarized the advantages and disadvantages of it.Applying the neural network control mode to control the lower limb exoskeleton and simulated in MATLAB.Then,compared the control effect of traditional control mode and neural network control mode.The results showed the neural network control arithmetic improved the performance of whole control system.Furthermore,the gait data of normal subjects walking in different gradients of slope was imported to the neural network control system that had built,then,simulated the training results which could verify the rationality of the design of control system.