In this paper, in the presence of unknown actuator faults, input saturations, and complete unknowns including both internal dynamics and external disturbances, exact trajectory-tracking problem of a surface vehicle (SV) is solved by creating a finite-time fault estimator based fault-tolerance control (FFE-FTC) scheme. By virtue of input saturations, smoothly saturated controls are separated from input nonlinearities including unknown faults, and thereby leaving faults-mixed unknowns be exactly observed by a finite-time fault estimator (FFE). By defining an integral sliding-mode (ISM) manifold and deriving affine controls with unknown gains from smooth saturations, Nussbaum technique is deployed to synthesize a uniformly adaptive finite-time controller working in a large range outside input saturations. Within the close range where saturations are removed, an ISM-based nonsmooth controller with finite-time auxiliary compensation dynamics is devised to finely stick tracking errors to the origin. Intuitively, large- and close-range control actions are triggered by measuring the ISM error, and thereby contributing to the entire FFE-FTC scheme which achieves exact fault-tolerance and unknown-rejection under input saturations. Lyapunov and nonsmooth syntheses prove that the closed-loop FFE-FTC system is globally finite-time stable. Simulation results and comparisons on a prototype SV demonstrate remarkable performance in terms of feasibly saturated controls and exact trajectory-tracking, simultaneously.

In the single-beacon underwater tracking system, vehicles rely on slant range measurements from an acoustic beacon to bound errors accumulated by dead reckoning. Ranges are usually obtained based on a presumed known effective sound velocity (ESV). Since the ESV is difficult to determine accurately, traditional methods suffer from large positioning error. By treating the unknown ESV as a state variable, a novel single-beacon tracking model (the so called “5-sv” model) and an extended Kalman filter (EKF)-based solution method have been discussed to solve the problem of ESV estimation. However, due to the uncertainty of underwater acoustic propagation, the probabilistic characteristics of the ESV uncertainty and acoustic measurement noise are unknown and varying both with time and location. EKF, which runs with presupposed noise parameters, cannot describe the practical noise specifications. To overcome the divergence issue of EKF-based single-beacon tracking methods, this paper proposes an adaptive Kalman filter-based single-beacon tracking algorithm which employs the “5-sv” model as the baseline model. Through numerical examples using simulated and field data, both the filter and smoother results show that while implementing the proposed algorithm, the tracking accuracy can be significantly improved, and the estimated noise parameter agrees well with its true value.

In recent years, with the use of the float-based wireless sensor network in the Argo project, a large amount of ocean data has been obtained. These data can be applied to analyze the oceanic thermocline, and the forecasting trend of the thermocline through the SVR method in machine learning is presented in this paper. Firstly, this paper refines the spatial resolution with the SVR method and determines the lateral boundary of the three-dimensional thermocline through the information entropy method. Combined with BOA Argo data from 2004 to 2015, this paper then predicts the thermocline trend (10°–25° S and 55°–80° E) over the next 4 years. The results show that the trend of the three-dimensional thermocline’s lateral boundary can be effectively predicted with the application of SVR method.