The Dynamic Neural Fuzzy Model-Based Motion Control For Underactuated Surface Vessels

It is so difficult to control the three DOFs of the horizontal positions and the course of the Underactuated Surface Vessel (USV) simultaneously, only depending on the two control inputs which are the fore and aft propulsion generated by the propeller and the turning moment generated by the rudder. Moreover, because of ship dynamic possesses with great inertia, big time-lag and high nonlinear, and the modeling parameters perturbing, the environment disturbances such as wind, wave and current, and the imprecise measurements, the problem of USV motion control exists inherent uncertainties. The conventional linear or nonlinear control algorithms are always difficult to acquire integrated optimizational control performances. So it's very important and essential to discuss some type of new intelligent control algorithms for USV.A Dynamic Neural Fuzzy Model (DNFM) is presented firstly for identifying the ship inverse dynamics. The DNFM adjusts its structure and parameters simultaneously when learning. And it can carve up the input space and decide the numbers of membership functions and fuzzy rules automatically, without any transcendental expert knowledge.Secondly, the DNFM is combined with a classical PID controller to construct a new controller named DNFPIDC for uncertain ship course control, which has the two modes of off-line and on-line learning. The simulation results indicate that the DNFM is suitable for course control because of the fast learning rate and well identification.Thirdly, the design method is extended to solve the problem of USV motion control. Two DNFMs and a Nonlinear Controller (NC) are combined to construct a trajectory-tracking control algorithm named DNFNC. And the convergence prove of the tracking errors is given following. The simulation results validate the feasibility.Finally, the DNFM is extended to a generalization DNFM named GDNFM. And an adaptive control algorithm named DNFAC is proposed based on the GDNFM for the trajectory-tracking of USV as well, and also giving the stability prove of the control system and the simulation validation. The comparative explanations of the above mentioned DNFM-based control algorithms are given in the Chapter 6.The research outcomes of the DNFM-based USV motion control indicate that the DNFM presents fast learning and exact identification ability, and can conquer the uncertainties effectively. It is suitable for both the SISO problem of the ship course control and the MIMO problem of the USV motion control. The results in this dissertation provide a new method for solving the problem of USV motion control, which possesses some theoretical significance and practical values.