Research On Predictive Environment Modeling For Space Tele-Robot

Space tele-robot with human-robot interaction is a powerful tool for performing tasks in unknown or hazard environment. Due to the large communication time delay between the tele-robots in space and the operator at the local site, the stability and maneuverability of the system are decreased. Control technology based on predictive environment model is an effective method to solve this problem. However, it depends heavely on the accuracy of the virtual environment model, and the negative effects increase rapidly as the time delay becomes larger.In this dissertation, a novel method is proposed to precisely model the environment of the tele-robot. The properties of the graphic and dynamic model of the space teleoperation and the relation between them are analysised. The virtual environment model is built at the local site based on the visual and force information from the remote side. The interaction between the slave arm and the environment is divided into three pecedures: transition to and from contact conditions and stable contact. Different models are used to describe the procedures respectively. In order to improve the frequency of the virtual force feedback, the force calculation and graphic update are done in two separate loops. Graphic model is overlayed on the video images by a novel register method to observe the results of the commands both in the virtual and real environment and to correct the model when the discrepancy is beyond a certain threshold. Sliding average least square method is proposed to identify the dynamical parameters of the real environment with position and force feedback and the results are used to update the corresponding parameters of the model.Besides the theoretic analysis, considering of the small gravity in space, a tele-robot system is developed, which is in fact a one-DOF master-slave teleloperation system with visual and force feedback. The proposed method of modeling and modifying on-line is verified in this system by experiments. Results showed that the technology proposed in this dissertation reduced the effect of time delay in teleoperation and that with time delay of 0~30 seconds the errors of virtual force reflection are less than 5%F.S. after model modification.The research in this dissertation not only advances the theory in virtual environment technology, but also solves the problems of constant time delay in a certain range. It is promising that teleoperation system based on virtual predictive environment will improve the performance of space robot to fulfill tasks such as science experiment, daily maintenance in space station and repair of the on-orbit satellite.