Research On Hand-eye Coordination System Based On Humanoid Behavior Intelligent Robot

With the continuous integration of information technology and industrial technology,the robot-related intelligent industry is developing rapidly.The field of use of robots has moved from the traditional fields of mechanical parts manufacturing and automobile manufacturing assembly to emerging fields such as intelligent logistics and intelligent manufacturing.The traditional automated production environment is generally a structured environment where the placement of objects is determined.At present,with the rise of industries and the rise of artificial intelligence and other disciplines,the requirements of robots to grab objects are becoming more and more“personalized”and the scenes are complicated.This has become an urgent problem to be solved.This topic explores and studies the intelligent hand-eye coordination system.This topic uses the RealSense SR300 RGB-D camera as a visual sensor and builds a smart hand-eye collaborative system operating platform with the computer and the Mitsubishi robotic arm RV-4F-1D-S 11.The research done in this topic is mainly about the complete closed-loop development of the intelligent hand-eye coordination system using C++language,mainly in the following aspects:[1]Obtain scene RGB map,depth map and point cloud map through Intel RealSense SR300 RGB-D camera,filter the comparison result of the acquired image and perform single target setting,double target setting and hand eye calibration to obtain relevant matrix parameters;[2]Using the local convex connection property to implement scene segmentation on the scene super-voxel connection and extracting the target object in the scene by combining the RSD feature values summarized in the OSD data set;[3]The optimal grab position of the object is studied,and a method for calculating the time-saving compared with the traditional grab method is proposed.[4]Forward kinematics modeling and inverse kinematics modeling of Mitsubishi robot RV-4F-1D-S11,in which inverse kinematics modeling uses innovative projection geometry;[5]Packaged the developed C++ program and used CUDA multi-thread acceleration to run the intelligent hand-eye coordination system developed by this project on the built hardware system and conduct the crawling experiment.