I. Introduction
As Euler–Lagrange (EL) systems are frequently encountered in engineering applications, for example, robotic systems [1], [2] and spacecraft [3], [4], the issue of automatic control of EL systems has attracted increasing attention from the control community [5]–[7]. Among various control goals for EL systems, the first and foremost one is to maintain safe system operation and, as such, it is essential for the control system to be fault-tolerant for actuation failures [7]–[10]. It is noted that most existing fault-tolerant control (FTC) methods can only cope with the partial loss of effectiveness (PLOE) faults [11]–[14], while the study of the total loss of effectiveness (TLOE) [15], [16] faults (dying power faults) is limited. Here, in this work, we consider the case of extreme actuation faults (TLOE faults) in which at least one actuator at some particular channel completely fails to work (either gradually or suddenly), which would cause fatal consequences if not addressed properly. For security-critical EL systems, FTC methods based on redundant actuators are normally utilized to account for TLOE failures. The FTC methods based on redundant actuators can be roughly classified into two types: dynamic redundancy (DR) [17], [18] and static redundancy (SR). The common solution in DR is to use a unit of fault detection and isolation (FDI) [19] or fault detection and diagnosis (FDD) [21], [22] to monitor each actuator closely such that when a total failure is detected, the signal for switching to a new standby actuator is triggered. However, the utilization of FDI or FDD usually causes time delay, which would further cause switching delays. It is worth mentioning that the FTC method based on SR [23] can allow two or more redundant actuators to operate in parallel, and if other actuators suffer from PLOE or TLOE failures, the remaining actuators can still work in coordination.