Studies On Intelligence Control Algorithm Of Trajectory Tracking Of Free-Floating Space Robot

The applications of the space robotics are becoming much broader and broader withthe continuous and deep exploration of the space. Space robotics will be used in helpingor replacing the astronauts to accomplish their space tasks, such as capturing, releasingand maitenancing the satellite, as well as assembling the station. In this way, spacerobots can not only reduce the dangerous situation and the labor intensity for theastronauts, increase the efficiency and quality of their work, but also save the cost ofspace groping. Therefore, space robotics, which has broad potential applications, playsan increasingly important role in space activities, and has been a research hot thatattracts close attentions from all over the world.Free-floating space robot, short for FFSR, is a kind of robot, which consists of thebody (i.e., the satellite) and the manipulators assembled on the body. It can float freelyin the space and assist or replace the astronauts in orbit services. In order to save boththe nonrenewable fuel and electricity and increase the orbit life-span of the satellite, thecarrier state controller and the base controller of FFSR are turned off when the robotsare working. In other words, the notable feature of FFSR is that it can float in spacefreely. If the micro-gravity in the space environment is neglected, there will be noexternal force affecting on the system. The potential energy and the momentum areconserved. Due to the lack of a fixed base for the FFSR, there are strong kinematics anddynamics couplings between the manipulators and the base. As a result, the movementsof the manipulators will affect the state of the base. Consequently, the problem of how tocontrol the FFSRs is much more complicated than that of the fix-based robots. Inaddition to the coupling problems, there are many uncertainties in the FFSR systems,including the parametric uncertainties, non-parametric uncertainties and some other hardto be forecasted uncertainties. So, in order to track the desired trajectory quickly andaccurately, the designed controller must have the adaptive and anti-interferencecapabilities, which can update the control parameters according to the managed objectsautomatically and limit the interferers.Considering the uncertainties of the system, this dissertation studies the kinematicsand dynamics modeling problems of the FFSR and discusses the trajectory trackingproblem of the end-effector in joint space and task space. The concrete research worksand the primary achievements are as the following:Firstly, the kinematics and dynamics modelings of FFSR are the theoretical basisfor the intelligence controller of the trajectory tracking. Under the assumption that no disturbance torque is applied on the system, the Generalized Jacobian Matrix, whichcould reflect the kinematics characteristics of FFSR, is derived in the inertial coordinatesystem according to the system geometry relationship. This dissertation shows that theGeneralized Jacobian Matrix is not only related with the geometic parameters, but alsorelated with the inertia parameters, which are time-varying and are hard to be accuratelyobtained. The dynamics singularity and the work space problems of the system areanalyzed. On this basis, the dynamics model of FFSR based on Lagrangian equation isestablished in the joint space. The characteristics of the dynamics model are introducedand the differences between the FFSR and the fix-based robot are compared.Secondly, the intelligence control problem of FFSR in joint space is discussed.After analyzing the uncertainties existing in the actual system, an adaptive controlmethod based on a fuzzy compensator is proposed in the presence of kinematic anddynamic modeling uncertainties. A fuzzy neural network controller, which combines thefuzzy logic reasoning and learning ability, is applied to approach the uncertainties in thejoint space. The controller has overall approaching ability, thus the external interferenceand uncertainties impacting on the system can be effectively restrained. By calculatingthe number of the fuzzy rules of the controller, it is found that the number of the fuzzyrules is increasing exponentially with the increase of the number of the manipulators. Inorder to reduce the amount of the calculation of the fuzzy rules and increase theoperational efficiency, an improved controller is designed. By analyzing and separatingthe uncertainties, the fuzzy rules are designed to approach the uncertainties, respectively.In this way, the number of the fuzzy rules is decreased significantly in comparison withthat of the controller previously designed. The controller is proved to be stable and theend-effector can track the desire trajectory rapidly.Thirdly, as a main factor affecting the control performance of the power drivesystem, dead zone is one of the essential research focus in the drive control research areawith high precision. In the joint space, an adaptive robust controller based on fuzzycompensator is designed to compensate the effect of the unknown dead-zone existing inthe FFSR system. A fuzzy compensator and the adaptive parameter updating law arepresented. Also, a robust adaptive controller is designed to adjust the output. Thestability of the designed system is improved. The simulation results show that the robustadaptive controller can inhibit the influence resulted from the dead-zone, and theend-effector can track the desire trajectory rapidly and accurately.Generally, the task of FFSR is given in the inertia space or the task space. In orderto realize the trajectory tracking of the end-effector in the task space, the adaptive control problem in the inertia space is discussed at last. The dynamic model of FFSR injoint space is mapped into the inertia space by augmented parameter method. Then, twoiterative learning controllers and the corresponding adaptive parameter updating lawsare designed and the stabilities of the systems are proved. The simulation results showthat after10times of iterative learning, the maximum tracking errors can be reduced toless than7%of those of the first circles, thus fast tracking in the task space is obtained.