| The intelligent bionic leg is also called the intelligent lower limb prostheses.The traditional passive or semi-active knee prostheses can not bring about a great change in the lives of amputees because of unnatural walking,great physical exertion and poor wearing experience of patients.The active bionic leg directly provides the active torque for the lower extremity joint through motor drive,and uses all kinds of sensor information to sense the motion intention of the patients and road condition,so as to better imitate the movement mode of the healthy leg of the human body,and greatly improve the life quality of amputees.It is of great practical significance for amputees to reintegrate into the society and lightening the burden of the society and the family.At present,in the man-machine hybrid system composed of active bionic leg and patient’s residual limb,there are still some problems: the lack of passive compliance of joint,the lack of stiffness adjustment according to the load change,the gait recognition problem under different tasks,the accuracy of motion intention recognition and gait prediction,and the joint motion control problem under the man-machine hybrid strong coupled system.In this paper,on the basis of elastic drive technology,a variable stiffness active knee joint and its bionic leg are designed and developed,and the bionic leg control platform is developed.Based on this platform,the gait phase recognition method based on biological information under different walking tasks is studied.At the same time,combined with physical sensors information,the motion intention recognition in more complex walking environment is studied.An adaptive robust controller based on gait trajectory is designed for force / position control of bionic legs.The main research work and innovations of this paper are summarized as follows:1.A variable stiffness actuator(Variable Stiffness Actuator,VSA)is designed for the knee joint of lower limb prostheses.On this basis,an active bionic leg,including virtual prototype and prototype,is developed.The stiffness regulation characteristics of elastic actuator are analyzed and calculated theoretically.The virtual prototype of bionic leg is established in the dynamic simulation analysis software ADAMS.The motion characteristics and energy consumption characteristics of bionic leg are analyzed by using the simulation model,and the effectiveness of elastic actuator in compliance control and energy storage is verified.The hardware control and software control system of bionic leg are developed,and the experimental platform of multi-source information acquisition and control is built.2.A gait phase recognition method based on surface EMG(s EMG)signal is proposed.A gait period is divided into six sub-phases: pre-swing,middle-swing,terminal-swing,pre-stance,middle-stance,terminal-stance.This provides important information for the phase control of bionic leg.In order to reduce the computational complexity,the time domain and frequency domain features of s EMG preprocessing are extracted.In order to effectively solve the problem of feature redundancy,the s EMG characteristics of key muscles of thigh are quantitatively evaluated and screened by using Davies-Bouldin Index(DBI)and Separability Index(SI).Then the Sequential Forward Selection(SFS)algorithm is selected to search the optimal feature set.In order to solve the problems of weak divisibility and recognition error increasement of feature samples under multitasking walking,a parallel Stacking integrated learning model is proposed.The optimal features are deeply excavated by using different base learning machines,which effectively improves the recognition accuracy and generalization ability of the algorithm under different walking tasks.3.A multi-level classifier fusion strategy based on stable gait recognition and transitional state recognition is proposed to identify the motion intention of human body,which can identify and predict five gaits: level walking,ramp ascent,ramp descent,stair ascent and stair descent.This provides key information for the motion switching of bionic leg.In stable gait recognition,acceleration,angular velocity,pressure and other physical sensors with s EMG signals are fused.The contribution of different sensor signals to gait recognition is analysed.The influence of different dimension reduction methods and different recognition algorithms on gait recognition is analyzed.The experimental results show that the recognition rate of the combined algorithm of Linear Discriminant Analysis(LDA)and Quadratic Discriminant Analysis(QDA)is 98.2%.In transition state recognition,a Light GBM(Light Gradient Boosting Machine)classification algorithm optimized by using the bayesian Treestructured Parzen Estimator(TPE)is proposed,and the synthesis optimal algorithm model of recognition accuracy and time are realized.Finally,a Hidden Markov Model(HMM)is proposed to combine the results of stable gait and transitional state to predict the motion intention of human body.4.A bionic leg control strategy based on gait trajectory is proposed.The joint motion trajectory is determined by perceived gait information.An adaptive robust force/position controller based on time delay estimation(TDE)is proposed to realize the bottom control of bionic leg.Aiming at the problems of nonlinear,uncertain and strong coupling of man-machine hybrid dynamic model,TDE technology is introduced,and an adaptive nonsingular fast terminal sliding mode control(ANFTSMC)is designed to realize model-free trajectory tracking control.In order to reduce the TDE error caused by external disturbance,a fuzzy neural network(FNN)compensator is designed.The co-simulation model of bionic leg is established in ADAMS/Simulink.The experimental results show that the proposed algorithm has obvious advantages under unknown interferences.By using the integral of time multiplied by the absolute value of the error(ITAE)and integral of the square value of the control input(ISV)indexes,the good adaptability of the algorithm under different stiffness of bionic leg is verified.Furthermore,the man-machine hybrid model is constructed in ADAMS/Simulink,and the natural walking of man-machine coordination is realized by using the FNN-ANFTSMC-TDE method proposed in this paper.At the same time,in the experiment of bionic leg prototype,the advantages of variable stiffness elastic driver in compliance control and energy storage are further verified. |
PDF Full Text Request