Research On Robot Teaching-playback Technology Based On Augmented Reality And Natural Human-robot Interaction

With the advent of intelligent manufacturing,robots have become an indispensable tool for modern industry.The current method of teaching-playback is still the preferred robot programming method for most small and medium enterprises.The increasingly diversified user needs and the increasingly individualized production model require that the robot teaching technology must keep pace with the changes.The augmented reality based teaching-playback,which combines the advantages of online programming and offline programming,has become one of the hot technologies in robot teaching-playback technology.In this paper,based on the research results of the robot teaching-playback technology and method in the early stage of the laboratory,a robot teaching-playback technique and method based on augmented reality and natural human-robot interaction technology are proposed.The use of augmented reality technology to register virtual robots with real robots coincides with them,enabling operators to quickly teach and verify virtual robots in real environments,while using natural human-robot interaction technology to reduce operator's cognitive pressure on robots.This method takes the GOOGOL GRB3016 robot as the object,performs DH modeling and its forward and inverse kinematics analysis,builds a virtual robot model on real scale and deploys it to the Holo Lens,adds motion control properties through Unity 3D,and applies the forward and inverse kinematics solutions for virtual robot motion control.The principle and steps of the superposition of the virtual and real robots are analyzed.The iterative method based on the orthogonal matrix is used to solve the homogeneous transformation matrix.The three-dimensional registration of the virtual robot is accomplished by Vuforia's multi-target tracking registration method.The gesture teaching-playback method uses Kinect to acquire the position and orientation of the operator's hand and controls the robot motion after the hybrid filter optimization.In order to improve the accuracy and stability of gesture teaching,Gaussian mixture regression model is used to predict the gesture position to solve the problem of gesture misrecognition.The speech teaching-playback method adopts Speech SDK to recognize the operator's voice instruction,and through the speech feature extraction and grammar file matching,the voice instruction is converted into a robot motion control instruction,and fine adjustment of position and posture is realized.The trajectory tracing experiment and the peg-and-hole experiment verify the feasibility and effectiveness of the research content in this paper.