Research On Virtual Environment Modeling Of Space Teleoperation System Based On Visual-Tactile Fusion

Due to the harsh and complex working environment in space,the human-in-loop teleoperation mode is the main working mode of space robots at present.In order to overcome the problem of time delay,virtual environment modeling technology is introduced into the field of space teleoperation.The virtual environment can provide the operator with visual,force and tactile feedback,so that the operator can observe the operation of the slave site in real time.However,space teleoperation system has high requirements on the accuracy of virtual environment,and the error of the virtual environment will cause the operator’s misoperation.For the task of space teleoperation,this thesis aims to build a teleoperation system based on virtual environment with high precision and good security.The research is mainly on geometry modeling of virtual environment for unknown object,dynamic modeling of virtual environment and the improvement of system security in teleoperation system.To meet the operational requirements of space teleoperation robots,a ground verification system for telerobot based on virtual environment is built.The kinematics of the system is analyzed,the forward and inverse kinematics modeling of the manipulator and the dexterous hand is realized,and the spatial mapping relationship between the hand controller and the robot is established.The slave robot is regarded as a known object,and the pre-defined model import method is used to complete the virtual environment geometry modeling of the slave robot.It provides a platform for subsequent research.Geometric modeling of unknown target based on visual information is realized through point cloud acquisition and preprocessing,point cloud registration combined with thick and thin registration,surface reconstruction and pose measurement.It is proved by experiments that the geometric model constructed by this method performs well in geometric dimension,positioning and attitude angle measurement.Then,according to the application background of space teleoperation,a reasonable target tactile point cloud acquisition and control algorithm is designed.Then the visual-tactile fusion method is used to fuse the visual point cloud and the tactile point cloud through KDtree,and the visual-tactile fusion point cloud is used to modify the geometric model.It is proved by experiments that the geometry model based on this method for objects with different materials is superior in geometric dimension,positioning and attitude angle measurement than the geometry model based on visual information only and that based on tactile information only.This method effectively combines the advantages of visual and tactile modal information.For dynamic modeling of virtual environment,because the object is floating in the space microgravity environment,it is impossible to carry out the task through direct contact.The dynamic modeling and parameter identification of space object are realized by visual-tactile fusion.Firstly,the material of the contact object is recognized by the facula image.While the data set of material facula image is established,the convolution neural network model for material recognition is built for training.Then,a more appropriate dynamic model and parameter identification method are selected for the contact objects of different materials through experiments.It is proved by experiments that this method can accurately identify materials,and the virtual force following errors of materials with different softness and hardness can meet the requirements of the system,and it also has a good force following effect in the presence of time delay.To address the safety issue caused by too fast operation of the operator,an adaptive speed control method based on operator recognition is adopted to control the manipulator.Firstly,the data set of operation time series is established according to the operation speed,and the CNNGRU fusion neural network model is built to recognize the operator.Then,according to the recognition results of the operator,the cascade PID controller is used to control the speed of the manipulator.It is proved by experiments that this method has high recognition accuracy for operators,and it can reduce the motion speed of the manipulator without changing the trajectory of the manipulator.The research in this thesis not only provides a high-precision virtual environment for space teleoperation system,but also helps to improve the system security,and lays a theoretical foundation for improving the efficiency and safety of space teleoperation.