ISA transactions最新文献

This paper addresses the data-based iterative learning control (ILC) problem for locally Lipschitz nonlinear systems, where the durations are iteration-dependent. A test framework is developed to perform test iterations for collecting specific input and output data from nonlinear ILC systems. By resorting to these data, an ILC updating law is provided through integrating modified outputs to compensate for the adverse effects of iteration-dependent durations. Thanks to the persistent full-learning property, a necessary and sufficient condition is proposed to accomplish the iteration-dependent perfect tracking objective, which depends on the output data. The developed ILC updating law that employs only data particularly applies to locally Lipschitz nonlinear ILC systems subject to irregular dynamics.

This paper investigates the formation tracking problem in collision-free multiagent systems (MASs). To address the disruption to control performance caused by external disturbances and unknown dynamics, an extended state observer (ESO) incorporating sliding mode and bias radial basis function neural network (RBFNN) is designed for expeditious estimation. The feedback from the observed information is then utilized to formulate a fixed-time formation strategy. Simultaneously, addressing these challenges increases the computational load and communication costs in MASs. Therefore, a distributed event-triggered mechanism is introduced to dynamically adjust controllers' update intervals. Furthermore, to address the high initial speed inherent in the fixed-time control strategy, a velocity-based artificial potential field (APF) is designed to prevent collisions between agents and alleviate actuator strain. The semi-globally ultimately fixed-time boundedness (SGUFTB) of the entire system is demonstrated via Lyapunov theory. The validity of the proposed strategy is subsequently confirmed through the execution of comparative simulation experiments involving five omnidirectional robots.

This paper focuses on flexible-joint robotic manipulators (FJRM), which possess numerous advantages such as high flexibility, precision, and fault-tolerance capabilities. However, FJRM are susceptible to various factors that may cause malfunctions during dynamic operations. These malfunctions not only compromise the operational stability and accuracy of the manipulator but also significantly shorten the equipment's service life. To address these issues, developing an effective control strategy is of significant practical importance. This paper proposes a novel adaptive performance enhancement (APE) control method to effectively tackle model uncertainties and actuator failures in FJRM systems. An adaptive neural network (ANN) algorithm is designed to achieve accurate trajectory tracking of uncertain robotic systems by compensating for modeling errors. A non-singular terminal sliding mode (NTSM) policy is proposed to realize compliance control of robotic manipulators, which enhances the system's robustness and interference suppression ability. The stability of the closed-loop system is subsequently validated using Lyapunov's direct method. Finally, the effectiveness of the proposed control method is demonstrated through simulations and experiments conducted on the Gazebo simulation platform and the Baxter robot. Comparative analysis with fuzzy neural network (FNN), neural network (NN) and PD control methods further underscores the superiority of the proposed method in terms of control performance.

This paper develops a torsional drill-string vibration suppression method based on model reference adaptive control and equivalent-input-disturbance considering high- and low-frequency disturbances. Most existing studies treat the bit-rock interaction merely as an undifferentiated external disturbance and achieve vibration suppression by compensating for it as a whole. However, they often fail to specifically address or decouple the distinct high-frequency and low-frequency components inherently present within the complex disturbance generated. To address this gap, this paper establishes a torsional vibration model of the drill-string. Then, a control scheme for torsional drill-string vibration based on the model reference adaptive control with equivalent-input-disturbance (MRAC-EID) method is developed. The state-dependent low-frequency bit-rock interaction disturbance is suppressed through the MRAC method, and the higher-frequency disturbance caused by the segment flexible characteristics of the drill-string is suppressed through the EID method. The stability analysis of the torsional drill-string vibration suppression system is conducted. Finally, the effectiveness of the developed MRAC-EID method is verified through simulation and micro-rig experiments, showing that the developed method achieves vibration suppression under different formations.

Robust online composition estimation is crucial for sustaining energy efficiency in industrial distillation processes, especially under dynamically drifting and unobserved conditions. Classical model-based estimators often encounter limitations in adapting to nonlinearities, signal drift, and measurement noise, while pure data-driven techniques fail to generalize to unseen trends and process drifts, resulting in suboptimal servo and regulatory performances. To overcome these limitations, this study proposes a hybrid data-driven soft sensor framework that integrates a Quadratic Program-based Constrained Kalman Estimator (QP-CKE) with Piecewise Linear Regression (PLR), along with an Adaptive Window (AW) extension. The AW-PLR dynamically adjusts regression window length based on a Cumulative Sum (CUSUM) derived nonlinearity index, enabling better sensitivity to transients and non-stationary process behaviour. The proposed soft sensors are validated for an ethanol-water mixture separation in a pilot-scale laboratory distillation column. Performance is benchmarked against Extended Kalman Estimator (EKE), static QP-CKE, and a Support Vector Machine (SVM)-based regression model. Quantitative evaluation shows that the proposed AW-PLR based QP-CKE achieves approximately 35 % lower RMSE, 22 % higher SNR, and 25 % faster computation time, while also offering noise resilience and process interpretability compared to its counterparts. The proposed hybrid soft sensors demonstrate enhanced adaptability, computational efficiency, and robustness, supporting their suitability for integration into real-time soft sensing and control frameworks.

Although the cyber-energy dual modulation (CEDM) has been applied to DC microgrids to achieve information exchange without additional communication lines, the communication bandwidth based on CEDM is very low. In this case, the traditional distributed collaborative control based on high communication bandwidth is not suitable. Thus, this paper proposes a distributed current edge-control strategy that fuses quantized state information with CEDM and combines dynamic event-triggered control with H_∞ consensus. A quantized CEDM protocol enables low-bandwidth communication while reducing communication costs. Simulations show output consensus under varied droop gains and external disturbances, with a maximum observed output error of 1.568×10^-2. The event-triggered mechanism lowers communication burden, yielding average triggering intervals ranging from 10^-2 to 10^-1 seconds.

This study develops an adaptive dual-terminal event-triggered control framework for secure affine formation maneuvering in multi-agent systems under false data injection (FDI) attacks. The framework integrates stress matrix-based affine localization with adaptive attack compensation, enabling followers to reconstruct attack-free states. A dual-channel triggering mechanism with time-dependent thresholds dynamically regulates communication based on state and input errors, balancing network load reduction with stability. Lyapunov analysis confirms bounded tracking errors and exclusion of Zeno behavior. Validation demonstrates effective attack mitigation, formation integrity preservation, robust performance, and communication efficiency.

Multirate sampled data (MRSD) dynamic systems are abundant in modern engineering systems. The inconsistent sampling rates cause data asynchrony and alter system properties, negatively impacting fault diagnosis with delayed and missed detections. In this paper, a fast rate fault detection scheme for dynamic systems is proposed, which is directly driven by MRSD. Firstly, the lifting technique is employed to transform asynchronous MRSD into single but slow rate sampled data. An auxiliary lifted output is constructed to compute a parity vector via subspace identification, facilitating a multi-dimensional diagnostic observer satisfying the Luenberger conditions. Then a post filter addresses causality constraint, allowing fast rate residual generation. Further, a fast rate residual evaluation scheme is developed. The effectiveness and superiority of the proposed scheme are demonstrated by a heating, ventilation and air conditioning (HVAC) example.

Distributed sliding mode controllers are proposed to address the optimal consensus problem for high-order nonlinear multi-agent systems under intermittent communication networks. Specifically, agents exchange information with neighbors only during non-overlapping time intervals, whereas all communication ceases completely during the interruption intervals. This feature significantly complicates the achievement of consensus. Under partial observability constraints, unmeasurable states are estimated via adaptive state observers. Additionally, a distributed optimization algorithm is employed to minimize the cost function and construct the optimal reference signal. To mitigate the problem of non-existent high-order derivatives, Hermite interpolation is adopted for optimal virtual signals. The distributed sliding mode controllers are designed to ensure that the tracking error of each agent converges to zero. Finally, stability analysis confirms the boundedness of the closed-loop distributed cooperative optimization framework, and simulation results verify the efficacy of the proposed method in practical scenarios.

This work presents a prescribed performance backstepping sliding mode control framework designed for a class of nonlinear systems with unknown disturbances. A key feature of the proposed method is the incorporation of a dynamic-gain neural network observer to handle system uncertainties and estimate unmeasurable states. In contrast with static gain observers, it adaptively adjusts its gain in real time, eliminating precise tuning. Moreover, an innovative integral nonsingular fast terminal sliding mode control (INFTSMC) strategy integrated with prescribed performance control (PPC) is developed to ensure that tracking errors adhere to pre-specified transient and steady-state requirements, enhancing reliability in practical applications. The control method manages dynamics, while the neural network observer compensates for nonlinearities, ensuring robustness under uncertainty. The system stability is analyzed via the Lyapunov theory. Simulation results demonstrate the effectiveness of the method.