Observer‐Based Fixed‐Time Fault‐Tolerant Control for Constrained Euler–Lagrange Systems: An Adaptive Neural Approach

Abstract

This paper presents a robust fixed‐time fault‐tolerant control (FTC) framework for Euler–Lagrange (EL) systems subject to unknown dynamics, intermittent actuator faults, and asymmetric state constraints. To effectively handle mismatched uncertainties and actuator degradation without relying on exact model knowledge, this work proposes an adaptive neural backstepping sliding mode architecture. The framework is founded on three key structural components. First, a decoupled fixed‐time neural observer is developed to simultaneously reconstruct unmeasured states and estimate lumped disturbances. This observer structurally decouples the nonlinearity approximation from the linear estimation error injection, eliminating the estimation loop to enhance accuracy under severe faults. Second, a normalized error transformation is integrated with a time‐varying barrier Lyapunov function (BLF). This transformation enforces asymmetric constraints globally, ensuring safety without requiring a priori knowledge of the initial error sign. Third, adaptive radial basis function neural networks (RBFNNs) are utilized within the backstepping design to robustly approximate and cancel lumped uncertainties. To resolve chattering‐induced instability, discontinuous terms are rigorously smoothed prior to neural approximation. Lyapunov analysis confirms that all closed‐loop signals are semi‐globally uniformly ultimately bounded (SGUUB) within a fixed time. Comparative simulations validate the robustness and tracking performance of the proposed scheme under complex, multi‐stage fault scenarios.