Grasp Strategy And Simulation Of Manipulator Based On Neural Network And Gaussian Process

The robot system with multiple degrees of freedom is an important part of the humanoid robot,and it is an actuating mechanism to realize the dexterous operation of the robot.The robot arm can carry the manipulator to the designated workspace,and the manipulator can achieve to the dexterous operation.The combination of this macro/micro manipulator can solve many problems exist in the single degree of freedom gripper.At present,with the wider and wider application of robot arm and manipulator,the complexity of its operation environment and the degree of refinement required increasing.Accordingly,the requirements for manipulator grasping operations are also increasing.The grasp strategy of the manipulator is a very important research direction on the artificial intelligence.In this paper,in view of the dexterity and intellectualization of the manipulator,the grasping strategy is discussed and discussed.The specific research contents are as follows:(1)For movement between finger and object of the manipulator,the manipulator of kinematics and dynamics is discussed.The kinematics is mainly including position kinematics,and dynamics of robot hand includes finger link and base joint dynamics.Taking the single finger as the research object,the simulation experiment for the control method is carried out in MATLAB.(2)In view of the planning problem of grasping point,a grab planning method based on neural network and Gauss process is proposed.The mathematical model and the force closure of manipulator grasping are discussed.And the grasp mode planning based on neural network is analyzed,which the RBF network model in MATLAB is established training the grasping mode.A planner that used the selected feature as the input based on the Gauss process classification is designed so that the probability of the grabbing point is given.(3)In order to analyze the problem of grasping operation and contact in the strategy of robot hand grabbing,the virtual model of the manipulator/arm is established for the three kinds of basic shapes.According to the configuration of the grasping operation,the data of joint rotation angle and contact force were collected by data glove and force sensing gloves,and the data was imported into MATLAB for processing.After that,the simulation was carried out in ADAMS.