Development Of A Vision-based Service Robot Manipulator Grasping Object System

With the development of technology,service robots have been widely used in daily life.As the limbs of the robot,the robot arm plays an important role.Considering that the traditional robot arm usually moves according to the established process and cannot perceive external information,this system combines the robot arm with vision to improve the intelligence of the service robot.Firstly,clarify the overall scheme and hardware design ideas of the system.Explain the requirements of the system,clarify the overall structure,analyze the hardware of the system,determine the selection of each device,and give the drive and control modules of each part,and design the overall electrical hardware diagram of the system.Secondly,analyze and research the algorithm of the system.First,through Zhang's calibration method,the internal parameters of the D435 camera are clarified,and the hand-eye calibration method and external parameters are determined to realize the coordinate transformation.After analyzing the visual recognition method,using the LINEMOD algorithm,the color and depth information are combined for template matching to identify the target object and clarify the position information to achieve target positioning.Then according to the D-H parameter method,the structural parameters of the mechanical arm are determined,and its forward and inverse kinematics are analyzed.Finally,on the basis of the RRT algorithm,a two-way exploration tree method is used to improve efficiency,that is,the RRT-connect algorithm,to plan the motion of the robotic arm.Finally,design the software and complete the test and experiment of the system.First,the software design of the recognition algorithm is explained,and the target object is recognized after training the coke bottle.After that is the communication program design of the communication between the two controllers and the CAN bus.Then build a simulation platform,use Solid Works software to build a three-dimensional model,convert it into a URDF file,configure Move It! in the ROS platform,select KDL's kinematics solver,and select RRT-Connect in OMPL for motion planning.Finally,various parts of the test are carried out to complete the grasping of the target object by the vision-based mechanical arm.The grasping success rate is 82.5%.The reliability of the grasping system is verified through experimental analysis,and the purpose of the system is basically completed.