Mixture-of-Experts (MoE) is a sparse neural architecture that significantly increases model capacity while maintaining low computational complexity. However, deploying MoE-based large language models (LLMs) on memory-constrained edge devices remains challenging due to their substantial memory requirements. To address this issue, we propose FIRM-MoE, a fine-grained expert offloading framework designed to enable flexible and efficient MoE inference. The core insight of our approach is to reduce the risk of inaccurate expert loading by decomposing each expert into fine-grained sub-experts and then dynamically allocating them through a fine-grained scheduling strategy. To further reduce the error in expert loading, we introduce a multi-layer expert prediction mechanism and a resource-adaptive expert pre-loading algorithm to enable more robust expert allocation. This design allows our model to achieve more efficient expert utilization and improved resilience to prediction errors. We conduct extensive experiments to demonstrate the superiority of FIRM-MoE across diverse memory constraints. The results show that FIRM-MoE achieves up to 1.5× speedup and 2.8× memory savings in decoding, compared to state-of-the-art MoE offloading strategies.

Adapting pretrained Large Language Models (LLMs) for time series forecasting primarily relies on token-level linguistic-temporal alignment, leading to the stacking of logically disjointed tokens as input.While empirically effective, these methods overlook a fundamental capability of LLMs: modeling linguistic logic and structure, rather than merely processing token features.To address this limitation, we propose the Markovian-Guided Structure-Aware Alignment (MGSAA).Our core contribution is a framework that transcends pointwise feature matching to achieve global structural isomorphism between the linguistic and temporal domains.Specifically, MGSAA distills latent evolutionary patterns of language within LLMs into a Markovian state transition graph, which is transferred as a structural prior to the time series domain.Under this prior, time series patches are decoded into latent states and then aligned via state-constrained cross-attention.Ultimately, MGSAA generates a token sequence topologically isomorphic to the LLM's inherent mental structure, reactivating its reasoning capabilities for forecasting.Comprehensive evaluations across multiple benchmarks demonstrate that MGSAA achieves state-of-the-art performance, providing an innovative solution for cross-modal alignment in LLM for time series forecasting.

LoRa has emerged as a crucial technology that provides ubiquitous connectivity for geographically distributed Internet of Things devices. LoRa employs linear Chirp Spread Spectrum (CSS) modulation in its physical layer to enable long-range communication. Recently, nonlinear chirps have been proposed to replace linear chirps in CSS, to enhance the capacity and scalability of LoRa technology. However, the characteristics and advantages of nonlinear chirps are not yet fully understood. To unleash the potential of nonlinear CSS modulation, this paper presents a comprehensive theoretical and experimental analysis of the nonlinear chirp, focusing on key aspects such as noise resilience, error correction, curvature selection, and bandwidth expansion. Our findings reveal that nonlinear chirps are intrinsically more resilient to noise than linear chirps. Moreover, nonlinear chirps typically exhibit a single-place demodulation error under strong noise conditions, which has informed our development of a simple yet effective error correction design. Additionally, users should carefully manage nonlinear chirp curvatures and signal power compensation to avoid distortion while benefiting from the bandwidth expansion offered by nonlinear chirps. Building on these insights, we further refine nonlinear CSS modulation and implement our optimized design on US-RPs. Field experiments demonstrate that our design achieves significant performance improvements, marking a step forward in the practical application of nonlinear CSS modulation in LoRa networks.

Multi-tier computation offloading is crucial to address capacity constraints and improve flexibility for mobile devices. However, existing research on multi-layer computing offloading faces challenges like inefficient resource utilization and poor scalability, particularly in handling diverse computational tasks. To address these challenges, this paper proposes a distributed resource allocation and mixed task offloading framework for end-edge-cloud collaborative systems that support partial and full task offloading modes. First, we propose a three-tier network computing architecture and formulate a task-offloading utility maximization problem by jointly optimizing mixed task-offloading and resource allocation. The proposed problem is a mixed integer nonlinear program (MINLP), which we solve by decomposing it into two subproblemsresource allocationandtask offloading. Edge computing resources and bandwidth allocation can be independently optimized at each edge node with a fixed task offloading strategy. Cloud computing resource allocation, while convex, involves a global constraint, which we solve in a decentralized manner using a multi-agent optimization approach. Then, we propose a joint task offloading and resource allocation optimization algorithm, CNO-TORA, to obtain the solution to the formulated problem. The algorithm is supported by strong theoretical guarantees and is almost surely convergent to a globally optimal solution. Experimental results on a real dataset demonstrate that our algorithm is scalable to large-scale networks and outperforms baselines, achieving improvements in average system utility ranging from 4.01%-28.15%.

Collaborative learning enhances the performance and adaptability of multi-robot systems in complex tasks but faces significant challenges due to high communication overhead and data heterogeneity inherent in multi-robot tasks. To this end, we propose CoCoL, a Communication efficient decentralized Collaborative Learning method tailored for multi-robot systems with heterogeneous local datasets. Leveraging a mirror descent framework, CoCoL achieves remarkable communication efficiency with approximate Newton-type updates by capturing the similarity between objective functions of robots, and reduces computational costs through inexact sub-problem solutions. Furthermore, the integration of a gradient tracking scheme ensures its robustness against data heterogeneity. Experimental results on three representative multi-robot collaborative learning tasks show that the proposed CoCoL can significantly reduce both the number of communication rounds and total bandwidth consumption while maintaining state-of-the-art accuracy. These benefits are particularly evident in challenging scenarios involving non-IID (non-independent and identically distributed) data distribution, streaming data, and time-varying network topologies.

Condensing large datasets into smaller synthetic counterparts has demonstrated its promise for image classification. However, previous research has overlooked a crucial concern in image recognition: ensuring that models trained on condensed datasets are unbiased towards protected attributes (PA), such as gender and race. Our investigation reveals that dataset distillation fails to alleviate the unfairness towards minority groups within original datasets. Moreover, this bias typically worsens in the condensed datasets due to their smaller size. To bridge the research gap, we propose a novel fair dataset distillation (FDD) framework, namely FairDD, which can be seamlessly applied to diverse matching-based DD approaches (DDs), requiring no modifications to their original architectures. The key innovation of FairDD lies in synchronously matching synthetic datasets to PA-wise groups of original datasets, rather than indiscriminate alignment to the whole distributions in vanilla DDs, dominated by majority groups. This synchronized matching allows synthetic datasets to avoid collapsing into majority groups and bootstrap their balanced generation to all PA groups. Consequently, FairDD could effectively regularize vanilla DDs to favor biased generation toward minority groups while maintaining the accuracy of target attributes. Theoretical analyses and extensive experimental evaluations demonstrate that FairDD significantly improves fairness compared to vanilla DDs, with a promising trade-off between fairness and accuracy. Its consistent superiority across diverse DDs, spanning Distribution and Gradient Matching, establishes it as a versatile FDD approach.

This paper addresses the co-design problem of adaptive control for networked wind turbine systems that track the desired rotor speed while efficiently scheduling network communication. Unlike existing approaches, the proposed method integrates the control design and the communication considerations, ensuring asymptotic tracking of rotor speed under the generator torque saturation with significantly reduced communication load. Firstly, the communication scheme is developed using the dynamic event-triggering mechanism, which introduces the feedback signal in the sampling loop. Secondly, an auxiliary signal is designed to mitigate the negative effect of inevitable generator torque saturation, ensuring that the bounded controller asymptotically exits from saturation. Then, the adaptive torque controller is constructed with the compensation signals to guarantee asymptotic stability, and the measurement function for dynamic event-triggering is co-designed alongside the controller to regulate the sampling-induced error. Furthermore, the requirement for exact knowledge of the wind turbine systems is eliminated by utilizing an online approximator to learn the uncertain aerodynamics and parameters. Finally, it is theoretically proven that all the signals in the closed-loop system are bounded, and the rotor speed asymptotically tracks the reference rotor speed. The feasibility and advantages of the proposed method are demonstrated on the NREL 5-MW wind turbine using the high-fidelity OpenFAST simulation platform.

This paper addresses the event-triggered adaptive control problem with prescribed performance guarantees for a class of nonaffine uncertain nonlinear systems. Compared with existing works, intermittent communication in the sensor-to-controller channel for high-order nonaffine nonlinear systems is considered. Since the controller design involves intermittent variables arising from the event triggers in the sensor-to-controller channel and completely unknown nonaffine nonlinearities, traditional control schemes encounter significant challenges and usually render intricate controllers. To alleviate these complexities, a system transformation is involved to convert the system into an affine-like form, followed by a performance transformation independent of initial conditions, thus enabling the proposed method to encompass a wide range of such problems. Unlike traditional backstepping-based methods, a non-recursive control approach with an observer is employed in this work, which simplifies both controller and event-triggered mechanism design. It is theoretically proven that all signals in the closed-loop system are uniformly ultimately bounded, and the event-triggered scheme significantly reduces the communication burden while precluding the Zeno phenomenon. Simulation examples are provided to demonstrate the effectiveness of the proposed method. Note to Practitioners—This work is motivated by the control desiderata, including computational and communication resources efficiency, transient performance guarantees, and adaptability to the uncertain nonlinear, which emerge in the networked control systems, such as unmanned aerial vehicles, robot manipulators, and chemical processes. Although numerous works have been reported to solve part of these problems, there is no unified solution covering these control objectives. In this paper, we propose an event-triggered guaranteed performance control via intermittent feedback signals in a unified framework, which is applicable to a wide class of applications without controller redesign. The intermittent communication in the sensor-to-controller channel is co-designed with the adaptive controller, which provides a provable sampling scheme for the digital implementation of nonlinear networked control systems. The proposed method permits flexible performance specifications, including exponential and prescribed-time types, rather than theoretical asymptotic behavior, lending it readily to engineering purposes. Furthermore, the performance functions are designed to be independent of the initial error, eliminating the need for recalibration that hinders practical applications. The non-recursive controller with intermittent feedback and adaptive updates reduces both the communication and computation burdens, which is simple yet efficient, making our approach user-friendly and less demanding in practical applications.

Floating offshore wind turbines (FOWTs) hold significant potential for developing wind energy in deep-sea areas. However, they are prone to additional motions, heavy workloads, and undesirable fluctuations under combined wind-wave loads. To tackle these issues, this paper presents a novel adaptive blade pitch control strategy with guaranteed transient performance for FOWTs operating in above-rated wind speed regions. This strategy effectively suppresses platform pitch motion to enhance system stability while maintaining the power output at its rated value. Specifically, this method initiates by defining a filtered regulation error to solve the non-affine nature of its model. An adaptive blade pitch controller is subsequently proposed, integrating an online learning approximator to cope with the unknown dynamics caused by unmeasurable external environmental inputs and model uncertainties. To further mitigate approximation errors, a nonlinear robust control law with adaptive gains is employed, enhancing the robustness and adaptive capabilities of the controller. Superior to the traditional adaptive controllers, a significant advantage of this strategy is its ability to quantify and ensure regulatory performance within predefined constraints during both transient and steady-state stages, thereby achieving high-performance control in power regulation. Finally, simulation studies are conducted using the OpenFAST software to show the capability of the presented blade pitch controller, which guarantees stable power generation with guaranteed transient performance.

Floating wind turbines (FWTs) hold significant potential for the exploitation of offshore renewable energy resources. Nevertheless, prior to the construction of FWTs, it is imperative to tackle several critical challenges, especially the issue of performance degradation under combined wind and wave loads. This study initiates with the development of a simplified nonlinear dynamical model for a semi-submersible FWT. In particular, both the rotor dynamics and the finite rotations of the platform are considered in presented modeling approach, thereby effectively capturing the complex interplay between the platform, tower, nacelle, and rotor under combined wind and wave loads. Subsequently, based on the developed FWT model, a novel adaptive nonlinear pitch controller is formulated with the goal of striking a trade-off between regulating power generation and reducing platform motion. Notably, the proposed control strategy adopts a continuous control approach, strategically beneficial in circumventing the chattering phenomenon commonly associated with sliding mode control. Furthermore, the controller integrates an online approximator and a robust integral of the sign of the tracking error, facilitating real-time learning of system unknown dynamics while compensating for bounded disturbances. Finally, both the accuracy of the established nonlinear FWT model in predicting key dynamics and the superiority of the presented pitch controller are validated through comprehensive comparative studies. Note to Practitioners—This paper addresses the conflicting goals between power regulation and load mitigation for floating wind turbines (FWTs) to ensure the reliable operation of wind turbine systems. This remains an ongoing challenge due to the inherent complexity of existing FWT models, frequently resulting in controllers crafted using linearized representations that fail to accommodate real-world uncertainties effectively. Through the utilization of a simplified physical-based nonlinear FWT model, a novel adaptive nonlinear pitch controller emerges as a promising solution. Notably, the developed nonlinear FWT model elucidates the coupling between rotor and platform degrees of freedom clearly and succinctly, facilitating the design of intelligent controllers. Our approach demonstrates the capability to concurrently regulate power production and stabilize the platform. Additionally, an online approximator is integrated into the controller to capture system dynamics, thus augmenting adaptability and diminishing reliance on high-gain feedback compensation. Importantly, this control strategy holds promise for extension and implementation in various other renewable energy systems.

Blade pitch actuators for traditional onshore wind turbines primarily serve the purposes of regulating generated power to the rated value in high wind speed regions. Yet, when considering the floating offshore wind turbines (FOWTs) subject to wind and wave disturbances, it can be observed that improving power regulation often comes at the expense of exacerbating motion in the floating platform, leading to elevated platform loads. To address this issue, this paper proposes a novel robust nonlinear pitch controller specifically designed to achieve power production regulation while simultaneously mitigating platform pitch motion for FOWTs. Moreover, to tackle the challenge posed by unknown dynamics under wind-wave joint loads, the presented controller integrates a two-layer neural network (NN) for real-time learning of these unknown system dynamics. Meanwhile, a robust continuous term is introduced to alleviate the effects of residual reconstruction errors from the NN and external disturbances. Finally, the viability and efficacy of the proposed scheme are clearly demonstrated through comprehensive comparative studies with traditional pitch controllers conducted on the National Renewable Energy Laboratory (NREL) FAST platform.Note to Practitioners—This paper is motivated by the challenge of achieving the competing objectives between power regulation and load mitigation for floating offshore wind turbines (FOWTs) in high speed region. FOWTs have garnered significant interest in the field of renewable energy due to their advantages, including the ability to install high-powered wind turbines and reduced costs in deeper waters. However, there exist some limitations that existing FOWT controllers face including inefficiency in directly applying onshore WT control schemes, typically depend on precise turbine models and lack adaptability to potential uncertainties and errors encountered in practical situations. To tackle these limitations, this paper proposes a novel robust nonlinear controller for FOWTs considering the platform motion, which has the ability to simultaneously achieve power production regulation while platform stabilization. A two-layer neural network for real-time learning is developed to capture potential changes in system dynamics, while a robust continuous term is introduced to alleviate potential errors in the control scheme. This naturally provides adaptability and robustness against uncertainties while enhancing overall control performance. The effectiveness of the proposed control scheme is validated through simulation results.

For DC microgrids (MGs), real-time adjustment of current sharing ratios and secure voltage restoration are paramount for optimizing load allocation and enhancing dynamic performance. In this paper, a dual-objective distributed model-free adaptive control (MFAC) scheme is designed for the first time to guarantee voltage transient performance and adjustable current sharing. First, an output-constrained nonlinear MG model with ZIP (constant impedance, constant current and constant power) load is established, and subsequently it is converted into an equivalent unconstrained data model using system transformation and dynamic linearization techniques. Second, a new prescribed performance control algorithm with asymmetrical preset boundaries is proposed to restrict voltage transient responses. This algorithm is updated with real-time input and output data at discrete instants, making it independent of line resistance and ZIP load measurements. To enhance the robustness of the control method, an internal observer is designed to actively compensate for the unknown nonlinear dynamics generated by time-varying system parameters. The stability conditions of the transformed systems in the presence of ZIP loads and time-varying line resistance are derived, which can indirectly ensure the prescribed voltage performance of the original system. Finally, the effectiveness of the proposed control method is validated through some simulations and hardware experiments.

Traditional optimization control algorithms need to solve the Hamilton-Jacobi-Bellman (HJB) equation in real time, which requires a time-consuming training procedure and is difficult to apply in actual engineering facilities. To break this limitation, an inverse optimal adaptive neural pitch control scheme is first presented for variable-speed wind turbines (VSWTs) to achieve a comprehensive performance of power regulation and load mitigation in this paper. First of all, to facilitate the design of the controller, the non-affine VSWT model is converted into an affine model by employing the filter regulation error technology. Subsequently, the stability of the overall control system is demonstrated via the Lyapunov theory and the regulation error ultimately converges to an adjustable region near zero. Then, for a meaningful cost functional, the practical inverse optimality of the control system is realized. Finally, the OpenFAST (Fatigue, Aerodynamics, Structure and Turbulence) platform is utilized to compare the proposed scheme with existing pitch controllers on an NREL WP 1.5-MW three-blade VSWT to verify the feasibility of the presented scheme. Note to Practitioners—VSWTs are receiving increasing research attention due to their performance in maximizing power generation and mitigating effects of wind fluctuations. Under the turbulent wind condition, it has great practical significance to achieve stable power quality and alleviate fatigue loads on valuable components of VSWTs. How to improve the power quality and reduce the operation and maintenance cost of the turbine is a challenging problem. This paper presents an inverse optimal adaptive neural network control method for VSWTs under the high-speed turbulent wind condition.

The frequency and rate of change of frequency (ROCOF) signals acquired from phasor measurement units (PMUs) are instrumental in the security monitoring and feedback control of power grids, necessitating minimal response time and high precision. Existing algorithms rarely achieve a simultaneous balance between superior noise filtering performance and rapid response time when measuring frequency and ROCOF. To address this problem, this article introduces an innovative algorithm that integrates$\alpha \beta \gamma$series filters. The proposed algorithm leverages an$\alpha \alpha \beta$(aab) filter, which combines an$\alpha$filter and an$\alpha \beta$filter, and the reference P-class PMU algorithm from the IEC/IEEE 60255.118.1-2018 standard file. By providing an original mathematical modeling approach to aab parameter acquisition, this filter effectively amalgamates the advantages of the two foundational filters. Extensive experimental simulations demonstrate that our proposed algorithm has exceptional performance across various metrics of frequency and ROCOF measurement, achieving results that compare favorably with other advanced algorithms.

Deep learning methods have achieved state-of-the-art performance in industrial fault diagnosis within the supervised learning paradigm. However, annotated data are scarce in industry, which can lead to overfitting and hinder their application. To address this issue, this article proposes a semisupervised learning framework that leverages self-supervised learning on abundant unlabeled data and supervised learning on limited labeled data simultaneously. Self-supervised learning captures inherent evolutionary dynamics, while supervised learning focuses on discriminative features. Specifically, a cross-prediction task on two augmented views of unlabeled data is devised using contextual representation. These contextual representations are used to construct a relational graph of unlabeled samples, which is then aligned with the corresponding logits graph. By facilitating interactions between the two tasks, the proposed framework achieves efficient fault diagnosis. Experiments on the Tennessee Eastman process and three-phase flow Facility datasets demonstrate the superiority of the proposed framework over other label-efficient methods.

Time-series anomaly detection plays a crucial role in industrial fault detection. Most existing studies follow either an unsupervised setting, which is prone to false alarms, or a supervised setting, which is time-consuming and labor-intensive. To address these limitations, we adopt an innovative weakly supervised paradigm for industrial fault detection, where segment-level labels are provided during training, while point-level predictions are made during inference. Within this paradigm, we propose an innovative$C$-ary tree-based multi-instance learning (MIL) framework. First, the entire time series is represented as a$C$-ary tree, where nodes representing subsequences of different lengths are treated as instances in the MIL framework. This design allows for the detection of both point and collective anomalies. Second, to detect out-of-distribution (OOD) anomalies that are not visible during training, we develop a vector quantization module to memorize regular historical patterns. OOD anomalies are then detected when they show significant discrepancies from all memorized patterns. Finally, we enhance the MIL framework with an attention-based pooling mechanism that allocates greater focus on anomalous instances, further improving detection performance. To validate the effectiveness of our method, we conduct experiments on four real-world industrial time-series datasets. The results show that our method outperforms existing approaches by at least 6.01% in AUROC under weak supervision.

Due to the uncertainty of the wind-wave environment and the low damping characteristic of the in-plane direction, the lateral direction of monopile offshore wind turbines (MOWTs) frequently experiences significant structural loads, which will shorten the service life of MOWTs. To solve this issue, an optimal pitch control method based on adaptive dynamic programming (ADP) is proposed for MOWTs, which comprehensively considers both power generation and load mitigation performance. First, the motion equation of the tower together with monopile under wind-wave environment is constructed on the foundation of the Euler–Lagrange equation. For the convenience of the controller design, a novel nonlinear affine MOWT mathematical model is presented, and an ADP-based optimal pitch controller is proposed to solve the multiobjective optimal control problem. Then, the uniformly ultimately bounded (UUB) stability of the system is proved by the Lyapunov theory. Finally, to validate the feasibility of the proposed control method, a national renewable energy laboratory (NREL) 5-MW offshore code comparison collaborative (OC3) MOWT has been emulated using the OpenFAST platform developed by the NREL.

Dataset Distillation (DD) is a technique for synthesizing smaller, compressed datasets from large original datasets while retaining essential information to maintain efficacy. Efficient DD is a current research focus among scholars. Squeeze, Recover and Relabel (SRe2L) and Adversarial Prediction Matching (APM) are two advanced and efficient DD methods, yet their performance is moderate with lower volumes of distilled data. This paper proposes an ingenious improvement method, Distributed Boosting (DB), capable of significantly enhancing the performance of these two algorithms at low distillation volumes, leading to DB-SRe2L and DB-APM. Specifically, DB is divided into three stages: Distribute & Encapsulate, Distill, and Integrate & Mix-relabel. DB-SRe2L, compared to SRe2L, demonstrates performance improvements of 25.2%, 26.9%, and 26.2% on full 224×224 ImageNet-1k at Images Per Class (IPC) 10, CIFAR-10 at IPC 10, and CIFAR-10 at IPC 50, respectively. Meanwhile, DB-APM, in comparison to APM, exhibits performance enhancements of 21.2% and 20.9% on CIFAR-10 at IPC 10, CIFAR-100 at IPC 1, respectively. Additionally, we provide a theoretical proof of convergence for DB. To the best of our knowledge, DB is the first method suitable for distributed parallel computing scenarios.

LoRa has emerged as a crucial technology that provides ubiquitous connectivity for geographically distributed Internet of Things devices. LoRa employs linear Chirp Spread Spectrum (CSS) modulation in its physical layer to enable long-range communication. Recently, nonlinear chirp has been proposed to replace linear chirp in CSS, to enhance the capacity and scalability of LoRa technology. However, the characteristics and advantages of nonlinear chirps are not yet fully understood. To unleash the potential of nonlinear CSS modulation, this paper presents a comprehensive theoretical and experimental analysis of the nonlinear chirp, focusing on key aspects such as noise resilience, error correction, and bandwidth expansion. Our findings reveal that nonlinear chirps are intrinsically more resilient to noise than linear chirps. Moreover, nonlinear chirps typically exhibit a single-place demodulation error under strong noise conditions, which has informed our development of a simple yet effective error correction design. Additionally, users should carefully manage signal power compensation to avoid distortion while benefiting from the bandwidth expansion offered by nonlinear chirps. Building on these insights, we further refine nonlinear CSS modulation and implement our optimized design on USRPs. Field experiments demonstrate that our design achieves significant performance improvements, marking a step forward in the practical application of nonlinear CSS modulation in LoRa networks.

Zero-shot (ZS) 3D anomaly detection is a crucial yet unexplored field that addresses scenarios where target 3D training samples are unavailable due to practical concerns like privacy protection. This paper introduces PointAD, a novel approach that transfers the strong generalization capabilities of CLIP for recognizing 3D anomalies on unseen objects. PointAD provides a unified framework to comprehend 3D anomalies from both points and pixels. In this framework, PointAD renders 3D anomalies into multiple 2D renderings and projects them back into 3D space. To capture the generic anomaly semantics into PointAD, we propose hybrid representation learning that optimizes the learnable text prompts from 3D and 2D through auxiliary point clouds. The collaboration optimization between point and pixel representations jointly facilitates our model to grasp underlying 3D anomaly patterns, contributing to detecting and segmenting anomalies of unseen diverse 3D objects. Through the alignment of 3D and 2D space, our model can directly integrate RGB information, further enhancing the understanding of 3D anomalies in a plug-and-play manner. Extensive experiments show the superiority of PointAD in ZS 3D anomaly detection across diverse unseen objects.

Artificial neural networks (ANNs) have been widely used for tasks like electricity theft detection (ETD) in smart meters. However, due to the subtle mechanisms and inherent opaque characteristics, ANNs are vulnerable to attacks. Although this attack surface poses significant risks, it has been largely overlooked in industrial scenarios. To alert the widespread adoption of industrial intelligence, this article studies the impact of backdoor attacks in ETD for the first time and proposes a feature attention distillation (FAD) defense. First, the attack surface in current model training pipeline is analyzed, and the adversaries can embed malicious backdoors for specific triggers to escape ETD. Then, six prevalent ANN-based models are tested and the adversaries can bypass the backdoored ETD models with success rates over 90.53%, which would inevitably bring huge losses to electricity companies. We further argue that the electricity companies can mitigate such attacks when noticing the abnormal nontechnical loss. A novel FAD defense that aligns the intermediate feature maps between fine-tuned and backdoored models is proposed, which can eliminate backdoors more efficiently with few resources compared with two classic defenses. The average attack success rate can drop by 90.71% with slight impacts on ETD performance. This work sheds light on a novel but perilous attack surface, and raises a warning for the wide adoption of artificial intelligence in smart measurement scenarios.

To promote the efficiency and economy of wastewater treatment process (WWTP), a novel data-driven robust adaptive dynamic programming (RADP) algorithm is proposed to balance the control performance and energy consumption. Action neural network and critic neural network constitute the proposed method, both the control signal and system error are simultaneously considered as part of cost function for lower energy consumption and better guaranteed performance. Furthermore, a robust item is designed to suppress the unknown disturbances of WWTP system and environment. The introduced method requires no prior knowledge of WWTP, and continuously updates the control law with the input–output data from WWTP system via the least squares algorithm. Moreover, the Lyapunov theorem validates the stability of controlled system. The systematic simulations based on benchmark simulation model No. 1 are performed to verify the superiority of the proposed RADP method compared with other methods that can achieve a significant reduction in energy consumption of aeration and pumping while maintaining the control performance.

Data-driven fault diagnosis for critical industrial processes has exhibited promising potential with massive operating data from the supervisory control and data acquisition system. However, automatically extracting the complicated interactions between measurements and subtly integrating them with temporal evolutions have not been fully considered. Besides, with the increasing complexity of industrial processes, accurately locating fault roots is of tremendous significance. In this article, we propose an unsupervised spatial-temporal aware graph encoder–decoder (STAGED) model for industrial fault diagnosis. First, the high-dimensional measurements are constructed as a weighted graph to depict the complicated interactions. Then, the graph convolutional network, long short-term memory network and attention mechanism are applied to learn a comprehensive representation for multiseries. To enforce the model to better capture the temporal evolution, the dual decoder that performs reconstruction and prediction tasks simultaneously is adopted with a well-designed comprehensive loss function. By learning the spatial-temporal evolutions of datasets, faults can be diagnosed and located at a fine-grained level based on reconstruction deviations. To verify the performance of STAGED, experiments on the Cranfield three-phase flow facility and secure water treatment datasets are implemented and the results indicate that it can provide insight into fault evolution and accurately diagnose faults.

Anomaly detection in multivariate time series has been widely studied in one-class classification (OCC) setting. The training samples in this setting are assumed to be normal. In more practical situations, it is difficult to guarantee that all samples are normal. Meanwhile, preparing a completely clean training dataset is costly and laborious. Such a case may degrade the performance of OCC-based anomaly detection methods which fit the training distribution as the normal distribution. To overcome this limitation, in this paper, we propose MTGFlow, an unsupervised anomaly detection approach for Multivariate Time series anomaly detection via dynamic Graph and entity-aware normalizing Flow. MTGFlow first estimates the density of the entire training samples and then identifies anomalous instances based on the density of the test samples within the fitted distribution. This relies on a widely accepted assumption that anomalous instances exhibit more sparse densities than normal ones, with no reliance on the clean training dataset. However, it is intractable to directly estimate the density due to the complex dependencies among entities and their diverse inherent characteristics, not to mention detecting anomalies based on the estimated distribution. In order to address these problems, we utilize the graph structure learning model to learn interdependent and evolving relations among entities, which effectively captures the complex and accurate distribution patterns of multivariate time series. In addition, our approach incorporates the unique characteristics of individual entities by employing an entity-aware normalizing flow. This enables us to represent each entity as a parameterized normal distribution. Furthermore, considering that some entities present similar characteristics, we propose a cluster strategy that capitalizes on the commonalities of entities with similar characteristics, resulting in more precise and detailed density estimation. We refer to this cluster-aware extension as MTGFlow_cluster. Extensive experiments are conducted on six widely used benchmark datasets, in which MTGFlow and MTGFlow_cluster demonstrate their superior detection performance.

Infinite delays, especially mixed infinite delays (MIDs), always pose a great challenge for exponentially stability analysis of neural networks (NNs). In this article, we construct a new Lyapunov functional that contains an auxiliary function with the ability to compress infinite time delays to bounded ones, which can remove some of the previous assumptions on NNs systems. Then, several new sufficient conditions to guarantee the exponential stabilization of state-dependent switched inertial NNs with MIDs are derived under the designed adaptive intermittent controller. Finally, numerical simulations are provided to illustrate the validity of the obtained results.

Multivariate time series anomaly detection has been extensively studied under the one-class classification setting, where a training dataset with all normal instances is required. However, preparing such a dataset is very laborious since each single data instance should be fully guaranteed to be normal. It is, therefore, desired to explore multivariate time series anomaly detection methods based on the dataset without any label knowledge. In this paper, we propose MTGFlow, an unsupervised anomaly detection approach forMultivariate Time series anomaly detection via dynamic Graph and entityaware normalizing Flow, leaning only on a widely accepted hypothesis that abnormal instances exhibit sparse densities than the normal. However, the complex interdependencies among entities and the diverse inherent characteristics of each entity pose significant challenges to density estimation, let alone to detect anomalies based on the estimated possibility distribution. To tackle these problems, we propose to learn the mutual and dynamic relations among entities via a graph structure learning model, which helps to model the accurate distribution of multivariate time series. Moreover, taking account of distinct characteristics of the individual entities, an entity-aware normalizing flow is developed to describe each entity into a parameterized normal distribution, thereby producing fine-grained density estimation. Incorporating these two strategies, MTGFlow achieves superior anomaly detection performance. Experiments on five public datasets with seven baselines are conducted, MTGFlow outperforms the SOTA methods by up to 5.0 AUROC%.

The rapid development of unmanned aerial vehicle (UAV) market presents potential threats to public security and personal privacy, and the vision sensors are widely deployed to detect the invasive UAVs because of the intuitivity and accessibility of the video. However, the small pixel area and weak morphological characteristics of distant invasive UAVs pose a considerable challenge to the detection precision. Prior work on UAV detection simply focuses on the information fusion between different feature layers, but ignoring the feature information inside each layer. In addition, to detect small UAV in video streams, the motion information of the target is also a noteworthy feature. In this regard, we propose a feature super-resolution-based UAV detector with motion information extractor. The proposed network fully utilizes the motion information of UAVs between temporal frames and the spatial invariant features between different resolution frames to pursue a high-accuracy small UAV detection performance. Experiments on Drone vs Birds dataset are carried out, and it is demonstrated that a higher detection accuracy on small UAVs is achieved compared with the baseline.

For the existing adaptive constrained robotic control algorithms, the demanding "feasibility conditions" on virtual controller is normally inevitable and the extra limits on constraining functions have to be imposed, making the corresponding approaches more demanding and less user friendly in control development. Here, we develop a new neuroadaptive constrained control strategy for uncertain robotic manipulators in the presence of position and velocity constraints. First, a novel unified mapping function (UMF) is constructed so that the restriction on constraining boundaries is removed and more kinds of constraining forms can be handled. Second, by integrating the UMF-based coordinate transformation with the "universal" approximation characteristic of neural networks over some compact set, the developed neuroadaptive control completely obviates the complicated yet undesired "feasibility conditions." Furthermore, it is proven that all closed-loop signals are semiglobally bounded and the constraints are not violated. The effectiveness of the proposed control is validated via a two-link rigid robotic manipulator.

This article focuses on the event-triggered (ET) fuzzy tracking control for uncertain state constrained strict-feedback systems with unknown control directions. A novel nonlinear transformed function is developed to transform the constrained system states into the counterpart without any constraints. An ingenious adaptation law is developed to co-design the control law and the ET rule, thereby effectively compensating the sampling error caused by the ET rule under unknown control directions. Based on the presented adaptation law and the Nussbaum gain technique, a novel ET fuzzy tracking control strategy is proposed, which can handle the situations with and without state constraints in a unified way without readjusting the control scheme. Subsequently, the nonlinear transformed function is extended to the time-varying state constraints, and the corresponding ET fuzzy control scheme is also modified to guarantee the closed-loop boundedness. The proposed two control strategies guarantee the satisfactory tracking performance, avoid the violation of the prescribed state constraints, and decrease the communication load effectively. Finally, the usefulness of the developed methods is verified through two simulation examples.

Industrial Internet of Things (IIoT) is vulnerable to advanced persistent threat (APT). In this article, we study a scenario in which APT is launched to attack IIoT devices. Considering the APTs lateral movement, a node-level state evolution model is established to calculate the probability of every device in an IIoT system to be compromised by APT. Based on this, a Stackelberg game model is proposed for the APT attacker and defender, which can accurately describe the gaming process. An effective computational approach is developed to obtain the potential Stackelberg equilibrium strategy pair of the game. Extensive case studies and comparison studies are conducted to validate the effectiveness of the proposed method.

Recent approaches based on metric learning have achieved great progress in few-shot learning. However, most of them are limited to image-level representation manners, which fail to properly deal with the intra-class variations and spatial knowledge and thus produce undesirable performance. In this paper, we propose a Deep Bias Rectify Network (DBRN) to fully exploit the spatial information that exists in the structure of the feature representations. We first employ a bias rectify module, which is able to focus on the features that are more discriminative for classification by giving different weights, to alleviate the adverse impact caused by the intra-class variations. To make full use of the training data, we design a prototype augment mechanism that can make the prototypes generated from the support set to be more representative. To validate the effectiveness of our method, we conducted extensive experiments on various popular few-shot classification benchmarks and our methods can outperform state-of-the-art methods.

Intelligent transformation for traditional factories is a widely discussed topic. The key to this transformation is ensuring the integration between information technology and operational technology. However, it is a challenging task in industry owing to the communication heterogeneity of the underlying production equipment (horizontal communication), and inefficient interactions between the equipment and information decision center (vertical communication). In this paper, we explore asset administration shell (AAS), an asset virtualization technology, shielding heterogeneous physical communication protocol of production equipment. Besides, to promote inefficient communication between the equipment and information decision center, we adapt OPC UA protocol as the communication protocol of AAS for vertical communication. In addition, time-sensitive networking (TSN) is applied to ensure communication between the AAS and the corresponding physical device. Above operations ensure devices interconnection and interoperability. On this basis, we propose an AAS-based production management platform (AASPMP), which aims at the coverage from the demand side to the production side. Such an intelligent system characterizes three layers to decompose complicated system functionalities, and a visible client is provided for the convenience of remote operation and maintenance. We deploy our system on the actual production system and demonstrate the effectiveness of our design.

The tracking control of a wastewater treatment process (WWTP) is considered. The process is highly nonlinear, with strong coupling, difficult to model mathematically, and the operation is subject to unknown disturbances. We address this multivariable tracking control problem by applying the direct heuristic dynamic programming (dHDP)-based reinforcement learning control. The control goal is to track a desired reference of the dissolved oxygen (DO) concentration of the 5th aerobic zone ($S_{O5}$) and nitrate concentration of the 2nd anoxic zone ($S_{NO2}$) by manipulating the oxygen transfer coefficient of the 5th aerobic zone ($K_{L}a_{5}$) and internal recycle flow rate ($Q_{a}$). The dHDP aims at achieving a minimal accumulated WWTP tracking error while dealing with strong coupling between the$S_{O5}$and$S_{NO2}$and eliminating unknown disturbances in the process. The proposed dHDP approach devises an optimal control strategy entirely driven by WWTP process data as an online learning control method. We have conducted extensive and systematic simulations based on the well-known BSM1 platform of the WWTP controlled by dHDP to compare and contrast performances with other methods.

This paper addresses the synchronization control problem of leader-follower multiagent systems with each follower described by a class of high-order nonlinear multiple-input-multiple-output (MIMO) dynamics in the presence of time delays and actuator faults. A distributed synchronization scheme with guaranteed synchronization performance based on the radial basis function neural network (RBF NN) is introduced. We propose an augmented quadratic Lyapunov function by incorporating the lower bounds of control gain matrices and the actuator healthy indicator, and the problems caused by the unknown time-varying control gain matrices, actuator faults, and coupling terms among agents are solved. Meanwhile, the output of followers can track that of the leader and the steady state, and the transient performance of synchronization can be guaranteed, while all the other signals in the closed-loop system are guaranteed to be bounded. Finally, numerical analysis has been carried out to verify the effectiveness of the proposed controller.

For pure-feedback nonlinear systems under asymmetric output constraint, we present a low-cost neuroadaptive tracking control solution with salient features benefited from two design steps. In the first step, a novel output-dependent universal barrier function (ODUBF) is constructed such that not only the restrictive condition on constraining boundaries/functions is removed but also both constrained and unconstrained cases can be handled uniformly without the need for changing the control structure. In the second step, to reduce the computational burden caused by the neural network (NN)-based approximators, a single parameter estimator is developed so that the number of adaptive law is independent of the system order and the dimension of system parameters, making the control design inexpensive in computation. Furthermore, it is shown that all signals in the closed-loop system are semiglobally uniformly ultimately bounded, the tracking error converges to an adjustable neighborhood of the origin, and the violation of output constraint is prevented. The effectiveness of the proposed method can be validated via numerical simulation.

This paper deals with the synchronization control problem in the leader-follower format of a class of high-order nonaffine nonlinear multiagent systems under a directed communication protocol. A novel adaptive neural distributed synchronization scheme with guaranteed performance is proposed. The main contribution lies in the fact that both nonaffine agent dynamics, which basically makes most existing agent dynamics as special cases, and guaranteed synchronization performance are taken into account. The difficulty lies mainly in the nonaffine terms and coupling terms due to the interactions of agents. To overcome this challenge, an augmented quadratic Lyapunov function by incorporating the lower bounds of control gains is proposed. The problems resulting from the nonaffine dynamics and the coupling terms among agents are solved by incorporating the special property of radial basis function neural network into the derivative of the augmented quadratic Lyapunov function. The unknown nonaffine terms are addressed by using an indirected neural network approach. A nonlinear mapping is built to relate the local consensus error to a new one, which is subsequently stabilized via Lyapunov synthesis. As a result, the proposed approach can ensure the outputs of all follower agents to track the outputs of the leader, while the synchronization performance bounds can be quantified on both transient and steady-state stages. All other signals in the closed loop are ensured to be semiglobally, uniformly, and ultimately bounded. Finally, the effectiveness of the proposed controller is verified through a heterogeneous four-agent example.

This paper presents novel cooperative tracking control for a class of input-constrained multiagent systems with a dynamic leader. Each follower agent is described by a high-order nonlinear dynamics in strict feedback form with input constraints. Our main contribution lies in presenting a system transformation method that can convert the input-constrained state feedback cooperative tracking control of agents into an unconstrained output feedback control of agents with dynamics in Brunovsky normal form. As a result, the original problem is simplified to be a simple stabilization of the transformed system for the agents. Thus, the use of the backstepping scheme is obviated, and the synthesis and computation are extremely simplified. It is strictly proved that all follower agents can synchronize to the leader with bounded synchronization errors, and all other signals in the closed-loop system are semi-global uniformly ultimately bounded. Finally, numerical analysis is carried out to validate the theoretical results and demonstrate the effectiveness of the proposed approach.

In this paper, we present a novel adaptive consensus algorithm for a class of nonlinear multiagent systems with time-varying asymmetric state constraints. As such, our contribution is a step forward beyond the usual consensus stabilization result to show that the states of the agents remain within a user defined, time-varying bound. To prove our new results, the original multiagent system is transformed into a new one. Stabilization and consensus of transformed states are sufficient to ensure the consensus of the original networked agents without violating of the predefined asymmetric time-varying state constraints. A single neural network (NN), whose weights are tuned online, is used in our design to approximate the unknown functions in the agent's dynamics. To account for the NN approximation residual, reconstruction error, and external disturbances, a robust term is introduced into the approximating system equation. Additionally in our design, each agent only exchanges the information with its neighbor agents, and thus the proposed consensus algorithm is decentralized. The theoretical results are proved via Lyapunov synthesis. Finally, simulations are performed on a nonlinear multiagent system to illustrate the performance of our consensus design scheme.

An adaptive consensus algorithm is proposed for a class of nonlinear multiagent systems with completely unknown agent dynamics. Due to uncertainties in the agent's dynamics, previous consensus approaches usually yield uniformly ultimately bounded consensus error. Our main contribution includes a novel robust consensus algorithm which can guarantee that the consensus error converges to zero asymptotically. In order to address the unknown dynamics, a two-layer neural network (NN) is utilized to learn the unknown dynamics in an online manner, and a robust continuous term is introduced to alleviate effects of the NN residual reconstruction error and external disturbances. The continuousness of the control signal is guaranteed to remove the actuator bandwidth requirement and avoid the caused chattering phenomenon. The proposed consensus algorithm is distributed in the sense that each agent only exchanges information with its neighbor agents. The asymptotic consensus result is achieved via Lyapunov synthesis. Furthermore, the proposed algorithm can also be extended to the case where the agents are required to form a prescribed formation. Finally, simulation studies on a nonlinear multiagent system are provided to demonstrate the performance of the scheme.

In this paper, we present a novel tracking controller for a class of uncertain nonaffine systems with time-varying asymmetric output constraints. Firstly, the original nonaffine constrained (in the sense of the output signal) control system is transformed into a output-feedback control problem of an unconstrained affine system in normal form. As a result, stabilization of the transformed system is sufficient to ensure constraint satisfaction. It is subsequently shown that the output tracking is achieved without violation of the predefined asymmetric time-varying output constraints. Therefore, we are capable of quantifying the system performance bounds as functions of time on both transient and steady-state stages. Furthermore, the transformed system is linear with respect to a new input signal and the traditional backstepping scheme is avoided, which makes the synthesis extremely simplified. All the signals in the closed-loop system are proved to be semi-globally, uniformly, and ultimately bounded via Lyapunov synthesis. Finally, the simulation results are presented to illustrate the performance of the proposed controller.

In this paper, adaptive neural control is investigated for a class of unknown multiple-input multiple-output nonlinear systems with time-varying asymmetric output constraints. To ensure constraint satisfaction, we employ a system transformation technique to transform the original constrained (in the sense of the output restrictions) system into an equivalent unconstrained one, whose stability is sufficient to solve the output constraint problem. It is shown that output tracking is achieved without violation of the output constraint. More specifically, we can shape the system performance arbitrarily on transient and steady-state stages with the output evolving in predefined time-varying boundaries all the time. A single neural network, whose weights are tuned online, is used in our design to approximate the unknown functions in the system dynamics, while the singularity problem of the control coefficient matrix is avoided without assumption on the prior knowledge of control input's bound. All the signals in the closed-loop system are proved to be semiglobally uniformly ultimately bounded via Lyapunov synthesis. Finally, the merits of the proposed controller are verified in the simulation environment.