Fault-tolerant control of active compensation toward actuator faults: An autonomous underwater vehicle example

As the motions of faulty actuators may change the faulty influences and introduce uncertain losses, this paper focuses on developing actively compensation-based fault-tolerant control (FTC), where unpredictable influences from an actuator fault are taken as active inputs to compensate for the loss and the correspondingly optimal input appears as a root to be approached asymptotically. One-point secant-based FTC (OS) is proposed to provide limited input convergence where referential gradients are taken to construct intersecting lines for approximating the fault curve and approaching the root. Comparatively, two-point secant-based FTC (TS) is given to provide global input convergence while the gradient is generated by two consecutive fault points. Toward real systems where actuator inputs are limited to feasible regions, selective secant-based FTC (SS) is deduced on the basis of TS and SS, which generates the gradients by using previous fault points selectively. Since the convergent effects could be constrained in specific cases, selective parabola-secant-based FTC (SPS) is developed by constructing parabolic curves to approximate the fault curve selectively. Moreover, a conservative feasible region is determined for designing required inputs to ensure the availability of FTC inputs. Toward the cases where the faulty actuator collaborates with fault-free ones in finite dimensions, SPS is extended to generate multidimensional FTC inputs. Comparative simulations against the rudder fault of an autonomous underwater vehicle (AUV) are carried out to illustrate the effectiveness of the proposed methods.