Showing papers by "Mingming Wang published in 2015"

Constrained Model Predictive Control for Real-Time Tele-Operation Motion Planning

Abstract: The increasing demands of On-Orbit Servicing (OOS) require new technologies to complete the OOS mission. Space robot, since it is flexible, multi-functional and extendable, is one of the most promising solutions for OOS. In this thesis, a distributed real-time simulation architecture based on DDS has been presented for space robotic tele-operation tasks. Besides, a constrained MPC framework considering system input/output, anti-collision and anti-singularity constraints has been developed for the free-floating space robot. The objective of this thesis is to provide a general simulation architecture with intuitive perception for the operators on ground, and a new control framework considering multiple constraints for the application of the space robot while the performance and effectiveness are improved. For that purpose, a new distributed real-time simulation architecture, RACOON, has been implemented based on DDS in the environment Matlab/Simulink/Stateflow. The new simulation architecture provides the operator on ground an intuitive view of space robot and makes the simulation architecture open for collaborative tele-operation. As a complete simulation system, the multi-body dynamics, AMM, path & trajectory planning, motion control, virtual reality etc. subsystems, integrate together seamlessly to complete the whole space robotic missions. In order to realize the new control framework based on NMPC, firstly, the dynamics of the space robot with tree structure using the concepts of graph theory and spatial notation is introduced. A new IMM is derived from the topology of the space robot, which can be employed to analyse the sparsity of the JSIM and the complexity of the CRBA algorithm, and assist the decomposition of the JSIM. Secondly, the singularity and collision avoidance issues of redundant manipulators are investigated considering the redundancy and task priority. A so-called STR method is developed based on the concept of manipulability ellipsoid for singularity avoidance. For collision avoidance, a new strategy combining two control points is proposed to restrain the vibration of joint velocity and generate smoother joint trajectory reference. Thirdly, application of NMPC to space manipulator in capturing an uncooperative target satellite is investigated. The system input/output, collision/singularity constraints in practice imposed on decision variables are translated into linear inequalities as part of NMPC. An on-line QP algorithm with prioritized constraints is adopted to find the optimal control efforts. The effectiveness and performance of the new proposed methods in this thesis are demonstrated by comparison to traditional methods. Well-designed end-effector path, together with the NMPC guarantees the success of the space manipulator to complete the capture of the un-cooperative target satellite. This work shows the feasibility and validity of constrained MPC applied in the field of space robot. Furthermore, it can be used to support further OOS mission and space exploration.