Research On Grasping Control For SMA Bionic Manipulator

With the development of robot technology,the manipulator,as the end effector of robot,has made great progress.Most of the conventional end manipulator is powered by the traditional motor,but such drive structures inherently suffer from poor compliance,low power energy density,and high mass and noise.However,the shape memory alloy(SMA)has outstanding energy characteristics and excellent driving sensing characteristics,which makes the robot further improve other characteristics of the manipulator while ensuring the joint motion torque and grasping load capacity remain unchanged.However,due to the nonlinearity of shape memory alloy and the diversity of manipulator driving structure,there are some deficiencies in motion control and visual perception.Therefore,based on the actual SMA bionic manipulator as the experimental platform,this thesis has carried out the following four aspects of research respectively:research on SISO model free adaptive control method,SMA manipulator wrist motion control research,SMA manipulator finger joint drive control research and SMA manipulator environment visual perception research.Research on SISO model free adaptive control method:Based on the model free control method designed for nonlinear system controller,different dynamic linearization methods and different linearization data models are used for nonlinear system,and the controller design scheme of the model free adaptive control algorithm based on compact format,partial format and full format dynamic linearization is given respectively;then the same time-varying control system is used to simulate the MATLAB example.Finally,the above three model free adaptive control algorithms are tested and verified by using the experimental platform of SMA bionic manipulator.Research on the motion control of SMA mechanical wrist joint:Based on the compact form dynamic linearization based mode free adaptive control algorithm,the input control rule function involved in the controller design is modified and improved to improve its response speed to signal tracking;and according to the actual expectation of SMA mechanical wrist joint is mostly constant,the(k-1)time and(k+1)time difference are introduced to further improve the adaptability of the controller to the actual experimental platform;at the same time,the stability analysis and proof of the improved tight form model-free adaptive control algorithm are carried out;finally,the rapidity and applicability of the improved controller are further verified by MATLAB simulation and SMA bionic wrist joint position control experiment.Research on the driving control of SMA mechanical finger joint:Based on the real-time online modeling of the model free adaptive control algorithm(MFAC),the concept of"acceleration" is introduced into its control law,and according to the time-varying characteristics of the finger joint force in the process of grasping,the "acceleration" item is generated by the k time and(k+1)time,so as to improve the ability of SMA finger joint to track the time-varying signal;at the same time,stability analysis and proof of the controller given by the SMA knuckle;finally through the MATLAB study And SMA bionic mechanical finger joint force control experiments to further verify the tracking performance and speed of the improved controller.Research on environmental visual perception of SMA manipulator:Based on color RGB(Red Green Blue)image to predict the three-dimensional attitude of the recognizable object in the field of vision of the actual camera.The YOLO-v2 high-level image feature extraction convolution network and the full convolutional neural network layer are used to complete the prediction of the coordinates of the two-dimensional projection points of the three-dimensional points in the world coordinate system in the image coordinate system.On the premise of obtaining the corresponding coordinates of the projection points in the image coordinate system,the rotation matrix R and the translation matrix T which from the world coordinate system to the camera coordinate system are obtained by using the PNP(Perpective N Point)algorithm.Finally,through the simulation of linemod data set and the prediction experiment of the actual grasping scene of SMA bionic manipulator,the actual performance of the attitude estimation model is further verified.For the above three parts of the work,firstly,the single degree of freedom tracking and driving control is carried out for the wrist and finger joints of SMA bionic manipulator,and the agreement between the control algorithm and the actual SMA manipulator is verified;secondly,the joint artificial grasping experiment is carried out for the wrist and finger joints of SMA bionic manipulator to verify the integrity and controllability of SMA bionic manipulator.Finally,the 3D visual pose estimation model is used to assist the SMA bionic robot to perform self-service unmanned operation experiments to verify the overall effectiveness of the underlying drive control and upper visual decision.