Research On Grasping Method Of Mobile Manipulator In Unstructured Environment

With the development of society and the advancement of science and technology,more and more researchers have begun to pay attention to the application of mobile manipulator in unstructured environments,such as field exploration,armed explosion protection,disaster relief,agricultural production,and household services.These environments are often accompanied by irregular movement spaces,irregular obstacles,restricted operations,and changeable situations,which bring difficulties and safety threats to the normal operation of workers and robots.In order to improve the working performance of the robot,researchers improve the environmental adaptability and grasping success rate of the robot by optimizing the structure of the mobile manipulator and the control program,and try to use the intelligent algorithm to improve the intelligence of the robot.The quality of path planning and the accuracy of target recognition and positioning determine whether the mobile manipulator can successfully complete the grasping task in an unstructured environment.To solve the problem of path planning and target recognition and positioning of the target grasping process of the mobile manipulator in the unstructured environment.The main research contents are as follows:(1)Taking the Rocr6 manipulator as the research object,the space kinematics model of the manipulator is studied.The advantages and disadvantages of the traditional D-H parameter method in the modeling process of the manipulator are analyzed,and an improved D-H parameter method is proposed to solve the problem of multiple sets of feasible solutions in the inverse kinematics solution of the manipulator,and the accuracy of the method is verified by the experimental deviation rate of 0.16%.(2)Taking the D435 depth camera as the research object,the target recognition and positioning method based on deep learning is studied.Based on the camera coordinate transformation matrix obtained from internal parameter calibration and external parameter calibration,by comparing the feature information of different color spaces and the evaluation indicators of convolutional neural networks with different network structures on the dataset,a high-precision comprehensive recognition accuracy of 97% is proposed.(3)In order to improve the efficiency and quality of obstacle avoidance planning for mobile manipulator,an ACO-RRT algorithm combining ACO algorithm and RRT algorithm was proposed.By constructing a framework for introducing different exploration utility functions and developing utility functions in different scenarios,the performance of the algorithm is improved.A simulation environment for obstacle avoidance planning of mobile manipulator was built,by comparing the path cost and global time-consuming of RRT,RRT* and ACO-RRT algorithms,the feasibility and superiority of ACO-RRT algorithm is verified.(4)Build an experimental platform for fruit grasping based on mobile manipulator,the whole operation process is divided into three parts: target identification and positioning,path obstacle avoidance planning and fixed-point dropping.It can be seen from the experimental results that the goodness of fit of the identification and positioning coordinates relative to the actual coordinates is 0.9964,which indicates that the identification and positioning system has high identification accuracy.Compared with the RRT algorithm and the RRT* algorithm,the global cost of the ACO-RRT algorithm is improved respectively 28% and 13%,reducing the global time consumption by 5% and30%,the feasibility and superiority of the ACO-RRT algorithm are proved.