Research On Robot Autonomous Grasping Operation For Scattered Objects

With the development of robotics and the increase in labor costs,more and more industries have begun to use robots for production activities.As a basic skill,the robot grasps the target object,which has very important value in various environments,and is widely used in household services,logistics sorting,object handling,workpiece loading and unloading and other fields.Autonomous grasping by a robot is a very complex task,involving a variety of cutting-edge technologies,such as target detection,pose estimation,motion planning,etc.At present,robots in industrial environments are less intelligent,and teaching methods are usually used.For grasping,or only have the recognition function for a single type of object,there is no good solution for grasping multiple types of objects scattered around,especially in the presence of stacking,occlusion,etc.In order to realize the autonomous grasping of scattered objects with occlusion,stacking,etc.,the following researches are carried out in this thesis:(1)An object pose estimation method based on ORB-GMS is proposed.Aiming at the problems of low recognition accuracy and easy mismatching of traditional feature point detection algorithms,the ORB algorithm with fast running speed and high recognition accuracy is used for object feature detection,and the GMS algorithm is used to optimize.The experimental results show that,combined with the PNP algorithm,ORB-GMS achieves the pose estimation of the target object.(2)The target object pose estimation in the stacking case is realized.First,an instance segmentation network is built to solve the problem of target object recognition and positioning;then,through feature point detection and affine transformation,the object pose recognition problem is solved,and the principal component analysis method is introduced to improve the accuracy of target object recognition under stacking conditions.Rate.(3)The simulation environment is built,and the path planning method of robot automatic grasping based on obstacle avoidance is established and verified.Firstly,the RRT algorithm is improved by integrating the artificial potential field method to solve the problem of lack of target orientation;Then,according to the real environment of the robot grasping system,a simulation model of the working scene of the manipulator is built in ROS,and the obstacle avoidance planning of the manipulator is verified by experiments.(4)The experimental platform of robot automatic grasping system is built to verify the effectiveness of the algorithm.Firstly,the ROS-based control system node is written to build the software system,the hand-eye calibration of the robot system is carried out,and finally the grasping experiment is carried out on the robot grasping platform.The results show that the system built in this paper can effectively complete the task of grasping scattered objects under stacking conditions.