Research On Grasping And Impedance Control Of Dexterous Hand Based On Data-Glove

The study of Dexterous hands is one of hotspots in current robots technology research and one of key technologies of intelligent space robots. The operation of dexterous hands is very important for the space robot. Based on national high technology research and development program (863)"Research on new generation five-fingered anthropopathic dexterous robot hand and its cooperative control"(code: 2006AA04Z255 ), With HIT/DLR II dexterous hand for the platform, this dissertation improves the motion mapping based on data glove, improves the E-ANFIS model and using the Model to build the grasped model, and then proposes a flexible impedance control algorithm, finally, achieves the grasping control of the dexterous hands using the control system with data love.Based on human hand and HIT/DLR II dexterous hand, this dissertation establishes kinematics models of human hand and the dexterous hand, and analyses structural differences between the models. Then an improved fingertip position-based motion mapping is used to improve the mapping ability between the human hand and the dexterous hand, and experiment is made to prove that mapping algorithm is right or not. The results show that the position of the human hand is mapped to the dexterous hand very well. And it make sure that the veracity of grasping object.Although the mapping mentioned can prove that the dexterous hand can move with the human hand, because of the error of the sensors in the data glove, when the human hand is grasping the object, the dexterous hand can only get the rough station information by the mapping method. Considering the grasping task of dexterous hand with unknown objects, the model of grasped object is constructed by hyperelliptic equation. The improved E-ANFIS model can export shape and posture information of the grasped object. So that the veracity of the grasped object is assured. Information of the dexterous hand is used in the model. The newly grasped model is more accurate for the realization of multi-fingered manipulation. When the human hand is grasping the object, the station of the finger-tip is mapped to the dexterous hand, and its coordinate is used as the input of the intelligent model. And then the model of the grasped object is remade, and the exactly information of the object is obtained to prove the steadily grasp.Considering HIT/DLR II dexterous hand as the study object, a joint and Cartesian impedance controller with nonlinear compensation is designed. To improve the performance of the impedance controller, system parameter estimations with extended kalman filter and gravity compensation have been investigated on the robot hand.Finally, using the experimental platform of dexterous robot hand teleoperation system, Firstly, training data of the intelligent E-ANFIS model is gathered using the data glove, the match training to it is carried on. The capture experiment is made to prove the relationship between data glove and the grasped model. A lot of experiments are made to prove the effectiveness of the the control method. The results prove that the algorithm of the controller is correct.