Research On Technology Of Compliant Assembly Of Industrial Robot Fusing Vision And Force Sense

For a long time,parts assembly in the field of industrial production(especially some special equipment manufacturing industries)has a large amount of work,low efficiency,high labor intensity,and frequent manpower assembly errors,which seriously affects industrial production safety and corporate economic benefits.In recent years,with the progress and development of computer control technology and intelligent technology,industrial robots have gradually grown and matured,and have been successfully used in manufacturing related fields,laying a solid foundation for industrial assembly technology innovation and assembly technology intelligence.At present,technologies such as visual recognition positioning and force feedback control are commonly used in the assembly process of industrial robots.Although production efficiency has been improved to a certain extent,production costs have been reduced,but there are still low assembly accuracy,difficult positioning,large errors,easy wear and other issues.Enterprises urgently need to optimize assembly technology,further improve assembly quality,and enhance the competitiveness of their products.Based on the assembly experiment platform of a Research Institute,the following research has been done in this thesis:1)In order to realize the vision guided grabbing work of "eye in hand" fixed-point shooting in complex environment,an improved visual detection and positioning grabbing method based on yolov3 is proposed.The improved network is used to recognize and simply locate the assembly object in the image.The original image is clipped by the bounding box,and the image points are obtained by color segmentation and region division of the clipped image.Combined with the relationship obtained by the nine point calibration method,the grasping points in the robot coordinate system are calculated.2)In order to solve the problems of inaccuracy in one-time positioning,difficulty in obtaining image features and mutual influence between assembly parts,an image visual servo system based on convolution neural network is proposed.Seg Net is used to separate the parts to be assembled.According to the principle of image visual servo,the image motion feature conversion structure is built by modifying the Flow Net C structure and the motion offset direction of robot in Cartesian space is predicted.3)The force feedback control of the industrial robot is completed in view of the possible jamming phenomenon in the embedding process of the shaft hole assembly parts.Based on the principle of gravity compensation method,the real contact force is calculated,and the principle and shortcomings of impedance control are analyzed.A six degree of freedom machine human / position hybrid control is proposed.Through the selection matrix,the improved impedance control is combined with a priori to form a force / position hybrid control,and the embedding task is completed.