Research On Decoupling Technology Of AUV Based On State Feedback And General Predictive Control

The underwater motion controlling technology is the key and base of the intelligent research for underwater robot, in which the motion decoupling control technology plays an important part in research of underwater vehicle. It has significant meanings and practical values to raising the intelligent level of the underwater vehicle.On the basis of domestic and foreign references, related studies of nonlinear decoupling control technology is carried out in the paper, while taking the by the underwater robot-"HAIYI" controlled by rudders and thrusters as an model studied in respects of the three-degree dynamics model, coupling's characteristics and online identification。The causes to the existence of coupling in every degree of HAIYF"underwater robot is analyzed aiming at the dynamics model of the underwater robot controlled by rudders and thrusters; the dynamic equation of underwater robot is derived by Newton-Euler Equations; and dynamics model is simplified by virtues of the construction and motion characteristic of "HAIYI" in result of the dynamics model reflecting the coupling relations of horizontal, vertical and heading of "HAIYI" underwater robot. In the vertical dynamic model, based on ensured generality and reasonable model structure, the heading and transverse velocity has been reserved, whose algebra product is used as the coupling effect term between them. While in the heading and transverse models, the vertical and transverse velocity has been reserved respectively, and its algebra product is also used as the coupling effect term. Experiment is carried out to analyze and prove the characteristics of dynamics model coupling.Because of any uncertainty of information from both inner self state and external circumstance, the Kalman Fliter On-line Parameter Estimation Method is introduced directed against dynamics online identification. The method use recursive strategy to make parameter fitting, and allows parameter estimation in the situation of little data. Besides, this paper proposed an on-line switch method based on Lyapunov. This method is on the basis of online parameters of rotation models with switch condition as switch index to realized online model switch. The consistence reaching issues of stability and output errors has been studied theoretically and experimentally. The result shows the method is capable to improve the precision of online identification of model under interferes.In the research of underwater robot decoupling control technology, this paper introduces decoupling control algorithm combined with state feedback and generalized predictive control. A generalized predictive decoupling control method of model-switch state feedback is proposed with result of establishing the decoupling compensation network of the system. Thus, on-line and real-time solving state feedback and input transformation matrix is realized; the decoupling of the speed in direction of vertical and heading is achieved; the subsystem with single-input and single-output after decoupling is controlled by the generalized predictive control algorithms with concrete system performance requirements. On the basis of state feedback, the method further infuse the idea of rolling optimization with a lot of online information immersed into the control solution, improving anti-jamming capability of control system and inhibiting the amount of control fluctuations effectively. This method reduces the difficulty of decoupling control in nonlinear system improving the precision of speed control and dynamic response performance. Stimulation and experiment prove the method effective. Finally, the experiment result shows that model switch rate is proportional to the complexity of system tracking signals and performance of controller.