Disturbance observer-based composite neural learning path following control of underactuated ships subject to input saturation

This paper investigates the constrained waypoints-based path following control problem of underactuated ships in the presence of the actuator saturation and the unknown disturbance. An improved composite neural learning control algorithm is proposed by using the command filter and the robust neural damping techniques. In the proposed algorithm, the dynamic auxiliary system is established to generate the saturation error compensating (SEC) signal, which is used to modify the error dynamics such that the actuator saturation constraint is tackled. The neural networks are employed to deal with the model uncertainty, and the corresponding compensating effects are improved further by designing the simplified serial-parallel estimation model (SPEM). By constructing the robust neural damping term, only two adaptive parameters are required to be updated online. That leads to a smaller computational application burden. Furthermore, the composite disturbance observer (CDOB) is developed by fusion of the prediction error and the compensated tracking one, where the unknown disturbance can be estimated accurately and compensated effectively. In addition, considerable efforts are made to obtain the semiglobal uniformly ultimately bounded (SGUUB) stability of the closed-loop system. The convictive experiments are performed to verify the effectiveness and superiority of the proposed algorithm.