Research On Grasping Operation Of Omnidirectional Mobile Manipulator Based On Monocular Vision

As a compound robot,the mobile manipulator combines the characteristics of the mobile robot and the fixed robot arm,which not only gives the mobile robot more flexibility to complete a variety of tasks,but also greatly expands the working space of the robot arm to make better operating performance of the robot arm.This has good research and application value.In this paper,aimed to the basic task of robotic grabbing,an omnidirectional mobile manipulator system is constructed.Combining machine vision related technology,the robot grabbing operation is studied.An omnidirectional moving manipulator system was designed according to the requirements of the work task.The Mecanum wheel-based omnidirectional mobile platform was used,the structure of the Mecanum wheel was analyzed in detail.The roller busbar equation was derived based on the Mecanum wheel outer contour.According to the operation requirements,the components of the robotic grabbing system were selected.A kinematic model of the omnidirectional mobile manipulator system was established.The kinematic models of the Mecanum wheel omnidirectional mobile platform and industrial robot were established respectively,and the conversion relationship between the various coordinate systems of the mobile robot arm was established.The second-stage motion mode of the mobile platform-mechanical arm was proposed.The navigation scheme of the omnidirectional mobile platform was designed.The visual guidance method was selected and the path tracking control and path image recognition method were given.The target recognition and localization method based on monocular vision was studied.A pinhole camera model was established,The camera was calibrated with Zhang Zhengyou calibration method to correct the image,and then preprocess the corrected image.The feature point matching algorithm based on random tree classification was used to identify the target.Compared with the improved SIFT traditional feature point matching algorithm,the experimental results show that the random tree algorithm has good robustness and strong real-time performance.The experimental analysis of the monocular visual positioning show that the positioning is reliable.An omnidirectional mobile manipulator experimental system was built,and the hand-eye calibration was completed.Aiming at the problem of large error in the grasping operation,a error compensation method based on the position of the hand-eye was proposed.By correcting the position of the hand-eye,the error of the grasping is effectively reduced.The experiment of grasping and placing multiple groups of objects was completed.The experimental results show that the omnidirectional moving manipulator based on monocular vision has high precision and has certain application value.