Intelligent Neural Network Control Of Rigid-flexible Hybrid Free-floating Space Robots

Space robot systems will play an increasingly important role in the future spaceactivities. For the benefit of save control fuel, free-floating space robot system isconsidered. In order to improve operating performance, dual-arm space robot systemis necessary. Furthermore, the space robot system’s arm is usually designed to be lightand slender, so the flexibility of the arm can not be ignored.The dynamic equations of single-arm and dual-arm space robot systems areestablished through Lagrange equation of the second kind. Incorporating the assumedmodes method, the dynamic function of the space flexible manipulator is modeled.Based on mutual mapping neural network, two networks are used to computeforward kinematic and Jacobian matrix respectively, Lyapunov method is employed todesign convergent inverse kinematic control scheme for space robot.Combine fuzzy theory and wavelet based neural network, a fuzzy wavelet basedneural network control scheme is designed for dual-arm space robot system withunknown system parameters. Its network parameters are updating online by backpropagation algorithm so that fuzzy neural network has strong ability of self-learningand self-adapting.A fuzzy basis function network is adopted to approximate the unknown dynamicmodel of dual-arm space robot and a robust technique is applied to deal withuncertainties of modeling errors and external disturbances. The trajectory trackingcontrol schemes in joint space with entirely unknown system parameters and in taskspace with unknown system inertial parameters are proposed. The control schemesdon’t need to linearly parameterize the dynamic equations of the system. Meanwhile,all weights and parameters of the FBF network can be adaptively tuned online with alearning rule, and consequently the control scheme is flexible.Without active suppression of the flexible vibration in space flexible manipulatorsystem, a wavelet based fuzzy neural network control scheme is designed withunknown system parameters. In case of external disturbance is exist and systemparameters are unknown, a diagonal recurrent neural network is used to build theunknown inverse dynamic model and a control scheme based on diagonal recurrentneural network is designed. For the sake of actively suppress vibration of flexible link, based on singularperturbation theory, fuzzy logic control and vibration suppression optimum control forfree-floating space flexible manipulator with known parameters and partitioned neuralnetwork control and vibration suppression fuzzy control in the presence of unknownparameters are proposed. The two control schemes can not only dominate thetrajectory tracking but also damp out the vibration of the flexible link.Conception of virtual control force is applied to design hybrid trajectory whichintegrate both flexible mode and rigid motion, then a hybrid trajectory neural networkcompensating control scheme for space flexible manipulator with model uncertainty isproposed. The scheme not merely guarantees the robustness of neural network controlin the presence of model uncertainty, and that actively suppress vibration, so thetracking performance is improved.Numerical simulations illustrate the validity and feasibility of the proposedcontrol schemes.