Research On Parallel Robot Grasping Technology Based On 3D Point Cloud

With the concept of “Made in China 2025” and “Industry 5.0” of the European Commission,enterprises have demanding requirements for industrial manufacture automatization.The hot topic of future researches will be the intelligent grasping,sorting and assembly technology of industrial robots.First of all,the scene interaction ability of the robot is an important prerequisite for the completion of the task.The accuracy and efficiency of pose recognition and location grasping are the important manifestation the robot intelligence level.Secondly,some problems accompany the robot in the process of high-speed grasping,which are too much energy consumption,too much impact and so on.Therefore,the ideal terminal trajectory tracking is an important guarantee for the robot to work efficiently.However,the existing two-dimensional computer vision has large errors in position estimation and grasping point calculation and can only sort objects of fixed height,which makes it difficult for robots to complete the high-precision sorting and grasping tasks of multi-objects.Therefore,target positioning grasping method of parallel robot based on 3D point cloud data was designed in this paper.The main research work of this paper was expounded as follows.(1)The point cloud acquisition and preprocessing technologies were comprehensively analyzed.The point cloud data acquisition and preprocessing methods of these machines were introduced,which were point cloud online acquisition equipment,RGB-D camera,and object offline point cloud template acquisition equipment,handheld 3D scanner.The imaging principle and device calibration of 3D point cloud were analyzed.The pretreatment of point cloud included point cloud segmentation,filter denoising,point cloud simplification and so on.There are some problems in existing point cloud simplification algorithms.For example,key features are easy to be lost and complex potential surface information.In order to solve these problems,weighted local optimal projection(WLOP)point cloud simplification algorithm based on fast point feature histograms(FPFH)was proposed.With a simplification rate of 30%,a large number of key features could still be retained and enhanced the calculating speed of subsequent operation.(2)Aiming at the problem of target’s pose estimation and location of grasping point,an approach of parallel robot grasping technology based on 3D point cloud data was designed.The overall scheme of the robot 3D visual system was designed.The structural parameters and composition of the research object,Delta parallel robot,were introduced.Robot hand-eye calibration technology based on RGB-D camera was studied.The mathematical method of SVD decomposition was used to obtain the transformation matrix between the robot base coordinate system and the camera coordinate system,and the calibration error was 0.684 mm after calculation.Finally,the point cloud registration technology(RANSAC rude registration and ICP precise registration)was used to register the offline complete point cloud of the known centroid position and the online local point cloud.And then the hand-eye calibration matrix was used to convert it into the robot coordinate system and calculate the actual grasping points to guide the robot to complete the target grasping task.(3)In order to solve the problems of excessive energy consumption and high vibration and shock of parallel robot in the process of high-speed movement,this paper proposed multiobjective trajectory planning for parallel robot based on NSGA-Ⅱ.Geometric method was used to establish the inverse kinematic models for robot.This paper established trajectory mathematical model using interpolation based on quintic B-spline,and added kinematic constraints.This paper achieved multi-objective optimal of trajectory planning from time,consumption and impact aspects through NSGA-Ⅱ as well as achieved optimal time series.This method effectively reduced robot’s vibration and consumption in rapid running process.(4)The Delta parallel robot grasping experimental platform guided by 3D vision was built,and the designed localization grasping method of target objects was verified experimentally.Various objects were randomly placed in the robot workspace,and multiple grasping experiments were carried out.The experimental results showed that the robot grasping system based on 3D vision could precisely and stably complete localization and grasping of the target objects,meeting the actual production needs.