Research On Humanoid Lower Limb Control System Based On Pneumatic Artificial Muscles

Pneumatic Artificial Muscle(PAM)is pneumatic actuator which is made of intelligent material,similar to that of human muscle with compliance,meanwhile human limb is actuated by irregulately distributed muscles which can be classified as deep layers,media layers and superficial lays to achieve abduction-adduction,flexion-extension and rotation,so PAM has a promising application prospect in the limb bionic robot.In order to better understand muscle and joint biological properties throughout lower limb movement,mechanism structure and control system of humanoid lower limb based on PAM are designed first,which then simply and optimize the designed mechanical structure.Secondly,PAM charaters measured,kinematics and kinetic equation of hip joint are deduced.Finally,observer based neural network adaptive fuzzy control algorithm is proposed and verified by experiment later.The dissertation consists of six chapters,the main contents of this dissertation are as follows:In chapter 1,the development course of PAM and type of drive are described.PAM static Model,dynamic model and intelligent control strategies study state are reviewed both at home and abroad.Present state of pneumatic application in bionics is described.Classification of parallel manipulator,its application and control are introduced.Finally,research significance and contents of the project are summarized.In chapter 2,based on the basic principle of lower limb skeletal muscle,mechanical structural of multi-joint humanoid lower limb in the form of parallel manipulator driven by irregular distributed not the same length PAM and roll angle of ankle joint is transferred to knee joint,multiplex PWM waves by FPGA and drive plate for High Speed On-off Valve are designed.Critical components of PAM and High Speed On-off Valve are selected,building up control system of humanoid lower limb at last.In chapter 3,structure matrix of joint deduced,joint muscle forward and inverse kinematics is calculated with it.Simulation is proposed to evaluate displacement,velocity and acceleration characteristics of muscle and and the mechanical structure of humanoid lower limb is optimized.In chapter 4,Test-bed is constructed to test force output characteristic of PAM.Calculation of joint torque and moment of inertia identification are analyzed and then put forward kinetic equation in which PAM is simplified as single block actuators,combing with structure matrix of mechanism and virtual work,ultimately translates into state equation.In chapter 5,angular transducer feedback the pitch,yaw and angle as a whole instead of Pull transducer with each PAM,fuzzy adaptive is adopted to control hip joint as a whole,gradient descent algorithm of neural network to optimize and compute parameter of fuzzy control with Gauss basis function,robust control is proposed to guarantee robustness of control system,then observer is introduced to observe angular accelerations which cannot measured directly,adaptive laws is designed and stability of the control algorithm is verified with proposed Lypanov function.Finally,kinematic and kinetics properties of PAM and joint are analyzed by experiment in hip joint gait planning and tracking control.In chapter 6,approach,content,conclusion and innovation point of the research are summarized,and future prospect of the research is forecasted.