Studies On Neural Network Control Algorithm Of Trajectory Tracking Of Space Manipulators On Free-floating Base

With the development of technology in space, the explorations for space have been gradually extended. However, the harsh characteristics such as micro-gravity, high vacuum and intense radiation of space, leads high risks in space tasks, so the space manipulator attracts more attentions and plays an increasingly important role in space.Considering the uncertainties in spatial scale and ground scale situations, this dissertation studies the single-arm planar free-floating space manipulator and discusses the trajectory tracking problem in joint space.First, under the assumption that no disturbance torque is applied on the system, the kinematics model of the free-floating space manipulator system is derived base on the momentum conservation theorem. The dynamics model is established by Lagrangian equation. Then, the dynamics model of two-link planar manipulator system is derived, the motion characteristics are verified by simulation.Next, the uncertainties in the actual free-floating space manipulator system for space application are analyzed, and a neural network control algorithm is designed to solve this problem. The proposed controller can approximate the uncertain function by neural network, the function has been selected for the whole model and the uncertain part, and the robust term is used to overcome the approximation error. Then an adaptive law is designed to adjust the weight of neural network in real time on the basis of the stability of the system, thus restraining the influence of uncertainties and realizing the trajectory tracking of the joint angle. Comparing the simulation results of two controllers, it can be found that the tracking error for joint angle and angular velocity are smaller with the controller base on the uncertain part approximation, making it more accurate to track the desired trajectory in joint space.Finally, considering the motion situation transcale variation between space application state and ground experiment for the manipulator, the dynamics model of the two-link space manipulator under the influence of gravity in ground scale is derived. Then, the influence of gravity on tracking accuracy is analyzed by PD control algorithm, thus gravity compensation is introduced in the neural network controller which is suitable for spatial scale. Additionally, an error feedback law is used to enhance the anti-interference ability. The simulation results show that the controller based on gravity compensation is effective for the space manipulator when the transcale variation happens in motion situation.