The bolt joint is the key component connecting the rigid moving base body and the flexible manipulator. The dynamic characteristics of the flexible manipulator under the elastic constraint of the joint are analyzed, and the action mechanism of the elastic constraint of the bolt joint on the frequency and vibration mode is revealed. Considering the effects of line constraint and torsion constraint, the elastic constraint model of the joint is established. Based on the principle of virtual work, the boundary constraints of the joint end and the free end are established, and the analytical equation of frequency and the expression of vibration mode function are derived. The first three frequencies and vibration mode characteristics of the flexible manipulator under elastic constraints are analyzed numerically. The sensitivity method is used to analyze the effect of linear constraints and torsional constraints on the frequency, and the elastic constraint region is established to characterize the functional relationship between the binding stiffness and the natural frequency. It is found that under elastic constraints, the influence of torsional stiffness of bolt joint is mainly concentrated in the low-order modal frequency, while the linear stiffness has a great influence on each order modal frequency of the manipulator; With the decrease of elastic constraint stiffness, its influence on modal shapes gradually increases, especially on high-order modal shapes. The research results prove the internal mechanism of the influence of elastic constraints on vibration characteristics, which provide a theoretical basis for improving the dynamic characteristics of flexible manipulator.

Considering the control problems caused by uncertainties such as inaccurate modeling, external disturbance and joint flexibility, a neural network control method based on H∞ is proposed. By establishing the dynamic model of the free-floating space robot with flexible joints, according to its dynamic characteristics, it is split into a slow subsystem model representing the rigid characteristics and a fast subsystem model representing the flexible characteristics. Based on the H∞ robust control theory, a robust controller based on neural network is designed to realize the decoupling control of the rigid dynamic model, The designed weight adaptive learning rate can ensure the online and real-time adjustment of parameters. Based on Lyapunov theory, it is proved that the designed controller can ensure that the L2 gain of the system is less than the given index. A feedback controller based on velocity differential is designed to compensate the angle error caused by joint flexibility. The experimental simulation results verify that the proposed control method is effective and has good engineering application value.