TY - JOUR
T1 - Adaptive Neural Backstepping Control for Uncertain MIMO Systems With Sensor/Actuator Faults and Asymmetric Time-Interval Output Constraints
AU - Zhou, Ning
AU - Zhao, Canyang
AU - Cheng, Xiaodong
AU - Xia, Yuanqing
AU - Li, Tiejun
AU - Wang, Baohao
N1 - Publisher Copyright:
© 2013 IEEE.
PY - 2026
Y1 - 2026
N2 - This article deals with a class of strict-feedback multi-input–multi-output (MIMO) nonlinear systems, which are characterized by model uncertainties, unknown control coefficient matrices, nonvanishing external disturbances, and multiple unknown time-varying faults in both sensors and actuators. To achieve accurate reference trajectory tracking while fulfilling fixed time-interval asymmetric output constraints (FTAOCs), a novel filter-backstepping adaptive-neural-fault-tolerant constrained (FBAFC) algorithm is developed based on only partial information from faulty sensors. The controller integrates radial basis function neural networks (RBFNNs) to approximate unknown nonlinearities and compensate for compound uncertainties. A coordinate transformation is introduced for iterative backstepping design, and a new barrier Lyapunov function (BLF) with a shift function is derived in three phases. The convergence of tracking errors is guaranteed theoretically, even in the presence of multiple actuator faults. Finally, simulation results are presented in two scenarios to validate the proposed method.
AB - This article deals with a class of strict-feedback multi-input–multi-output (MIMO) nonlinear systems, which are characterized by model uncertainties, unknown control coefficient matrices, nonvanishing external disturbances, and multiple unknown time-varying faults in both sensors and actuators. To achieve accurate reference trajectory tracking while fulfilling fixed time-interval asymmetric output constraints (FTAOCs), a novel filter-backstepping adaptive-neural-fault-tolerant constrained (FBAFC) algorithm is developed based on only partial information from faulty sensors. The controller integrates radial basis function neural networks (RBFNNs) to approximate unknown nonlinearities and compensate for compound uncertainties. A coordinate transformation is introduced for iterative backstepping design, and a new barrier Lyapunov function (BLF) with a shift function is derived in three phases. The convergence of tracking errors is guaranteed theoretically, even in the presence of multiple actuator faults. Finally, simulation results are presented in two scenarios to validate the proposed method.
KW - Adaptation mechanism
KW - filter-backstepping adaptive neural constraint control
KW - multiple sensor and actuator faults
KW - strict-feedback nonlinear system
KW - time-interval asymmetric constraints
KW - unknown control gain matrix
UR - https://www.scopus.com/pages/publications/105027959715
U2 - 10.1109/TSMC.2025.3648683
DO - 10.1109/TSMC.2025.3648683
M3 - Article
AN - SCOPUS:105027959715
SN - 2168-2216
VL - 56
SP - 1761
EP - 1775
JO - IEEE Transactions on Systems, Man, and Cybernetics: Systems
JF - IEEE Transactions on Systems, Man, and Cybernetics: Systems
IS - 3
ER -