Recent efforts have yielded substantial progress in backstepping platoon control for connected and automated vehicles (CAVs). While most existing studies focus on guaranteeing individual vehicle stability and string stability, their deployment in nonlinear vehicle platoons may face challenges from the so-called "butterfly effect." That is, even with guaranteed string stability, potential instantaneous spacing changes may imply unpredictable, uncomfortable fluctuations in vehicular velocity and acceleration. To address this issue, a parallel error-fluctuation suppression control framework is proposed in this work. Specifically, tunable triple-layered error boundaries (i.e., spacing, velocity, and acceleration) are constructed to reactively confine all propagated errors within predefined envelopes. By integrating a Barbalat-lemma-enhanced filtering-compensating mechanism and an adaptive approach based on the approximation capability of radial basis function neural networks (RBFNNs), asymptotic error tracking is realized to proactively suppress potential fluctuations. An adaptive backstepping control approach-integrating proactive and reactive suppression strategies-is then proposed to mitigate the unquantifiable "butterfly effect." Theoretical analysis and simulations demonstrate the validity and superiority of the proposed approach.

Reliable bidirectional communication between the control center and manufacturing devices (MDs) along with efficient resource allocation are critical for the Industrial Internet of Things (IIoT). However, due to the limited network resources of IIoT devices and the intertwined nature of communication and computation resources, achieving efficient optimization of these resources presents significant challenges. In this article, we propose a multilayer communication architecture for the IIoT based on edge computing and cooperative communication technologies, where resource-constrained MDs upload data to edge servers (ESs) for processing. To address the issues of resource scarcity and coupling, we establish a bandwidth release model to analyze and quantify the relationship between computation and communication resources. This elucidates the interaction mechanisms between “transmission-computation” performance. Furthermore, to prevent excessive data upload to ESs, which could lead to node overload and network congestion, we employ game theory to develop a pricing mechanism for resource allocation, where ESs charge for computation services. Specifically, we construct a Lagrangian framework to determine a resource allocation scheme, aiming to maximize the utility of the factory. Simulation results indicate that compared to common methods employed in existing works, our strategy enhances factory utility by 29.79%.

In this paper, a resilient distributed model predictive control (DMPC) algorithm is proposed for the networked vehicle platoon system under denial-of-service (DoS) attacks. It is difficult to ensure the system-level stablibity and string stability of the networked vehicle platoon system simultiniously in the presence of DoS attacks. To this end, an optimization problem is established, which is related to the trajectories of the system. To minimize the cost function, the last-step shifting technique is employed and the assumed solution is determined by the local optimal solution. The last optimal solution without the occurrence of DoS attacks is used as the reference trajectory. This reference trajectory is then transmitted to the interconnected vehicles within the system. Demonstrating the stability of the closed-loop platoon system is achieved by using the Lyapunov function, which is designed by the sum of the cost function. String stability of the platoon system, the unique characteristic of platoon systems, is proven through rigorous mathematical derivation. Compared with existing results, the vehicle platoon system under DoS attacks can be converged both the internal and closed-loop stability. Finally, several simulation and experimental results are presented to verify the effectiveness of the proposed method.

In response to carbon peaking and carbon neutrality, DC microgrids ( MG), as a key pillar, have facilitated efficient and reliable power transmission between renewable energy sources, energy storage devices, and various loads. In the process, the heavy reliance on communication networks exposes them to potential cyber-physical security risks. Namely, attackers can inject false data to achieve current or voltage overload without triggering an alarm by eavesdropping the communication data between the converter and MG center. For this reason, an attack detection and localization framework using data-driven is constructed in this paper. Utilizing the subspace identification methods, a data-driven I/O model aiming at sketch the process input-output data-based framework for DC-MG dynamic processes is established. Afterward, the necessary theory on the data collected for the observability and controllability of the proposed data-driven model is given. Based on this, an attack detection and localization framework for data-driven design of DC-MG system is presented. The proposed framework includes a bank of adaptive residual generators, adaptive detection threshold and localization observers, whose parameters can directly be obtained from process data. Finally, simulation tests on the meshed DC-MG system consisting of four distributed generation units are presented to demonstrate the superiority of the developed attack detection and localization framework.

Underwater acoustic sensor networks (UASNs), which are formed by a number of interconnected mobile vehicles and static sensors, have emerged as a promising solution to explore and utilize the ocean resources. Typically, the localization, communication and control are the fundamental services for the applications of UASNs. Although they are closely related, the localization, communication and control issues are usually separately tackled. The separate design directly affects the localization accuracy, transmission reliability and control efficiency, especially for resource-constrained UASNs. In this regard, it is essential and necessary to co-design the localization, communication and control systems for UASNs. At present, the theoretical framework of the above integration design is still in the construction phase, and some key problems remain unresolved. Therefore, this article aims to give a survey on the integration design of localization, communication and control for UASNs. We first present the communication architecture, through which the main challenges aiming at the integration design are analyzed. After that, a holistic survey on the underwater localization, communication and control basics is provided. Followed by this, the recent advances on the integration design are given. At last, we make an outlook to the future research directions on the integration design of localization, communication and control for UASNs.

The high integration and continuous deformation capability of 4D-printed smart materials possess great potential in soft robot manufacturing. However, due to material properties and fabrication discrepancies, there may be variations in the sensing and actuation performance among 4D-printed actuators with identical structures, which can affect their synchronous operation. This paper develops a novel electrothermally driven 4D-printed multi-finger soft gripper system that integrates self-sensing, actuation, and control. The 4D-printed flexible structure integrated with flexible micro-crack sensor is used as the finger, and a simplified dynamic model is proposed based on the pseudo-rigid body modeling technique. In addition, an adaptive consensus control strategy is proposed to address the issue of performance response inconsistency among multiple fingers. The effectiveness of the proposed control strategy is verified by numerical simulation and experiments. Finally, bioinspired wrinkled structures made of silicone rubber are designed on fingers to improve the grasping performance of the gripper. The experimental results show that soft fingers with different initial states, driving, and sensing performances eventually converge to consistent states under the action of the controller even in the presence of disturbances. During the grasping process, the soft gripper demonstrated excellent coordinated deformation capability, effectively maintaining grasping stability.

4D printed constructs (4DPC) can transform into different forms and shapes upon exposure to environmental stimuli. However, most 4DPC can only achieve simple deformation under specific incentives currently. There is no effective strategy to enable the 4DPC to adapt to environmental changes autonomously. In this paper, an adaptive fuzzy controller based on expert knowledge is proposed, which will compensate for the disturbance and uncertainty caused by external stimuli to ensure the adaptability and robustness of the 4DPC in dynamic change and uncertainty. To facilitate the design of the controller, a long short-term memory (LSTM) neural network model is proposed to characterize the nonlinear deformation characteristics of the 4DPC. Finally, the environmental adaptability experiment of a self-made 4D printed color-changing structure(4DPCCS) shows that the 4DPCCS integrating perception and control functions can actively adapt to the dynamic changes of the environment. The deformation error of the regulation is$\le \pm 2.39\%$, and the 4DPCCS can adjust and maintain it adaptively according to changes in the external environment, achieving a camouflage function similar to that of a natural chameleon. Note to Practitioners—The aim of this study is to provide a new solution to meet the dynamic and real-time requirements of real environments for adaptive deformation of 4D printed constructions in dynamic environments. Most of the existing control methods act on the printing process of the constructions to guide the structural design of the 4DPC to achieve specific deformation to specific stimulus. This paper proposes a new method to realize the adaptive modulation of 4D printed constructions to external stimuli through the introduction of adaptive fuzzy control. First, we developed and designed a 4D printed color-changing structure with actuation, sensor and color-changing functions. Secondly, based on a large number of experimental data, we use the modeling method of the LSTM neural network to characterize the relationship between the complex deformation of 4D printed constructions and external stimuli. Finally, an adaptive fuzzy control system for 4D printed color-changing structure is developed and designed based on the neural network model. The experimental results show that this method is feasible. Under the action of the proposed control strategy, the constructions can actively identify the changes of the external environment and adjust and maintain adaptively, realizing the function of color-changing camouflage. In future studies, we will further research the autonomous adjustment method for variable performance or variable functionality of 4D printed constructions.

This paper addresses the fault-tolerantH∞output regulation problem of multi-agent systems (MASs) subject to input saturation, structural uncertainties, and actuator faults. TheH∞output regulation problem is reformulated as a distributed two-step zero-sum game problem to enhance both steady-state and transient performance in the presence of disturbances. A novel anti-saturation reinforcement learning algorithm with a feedforward-feedback structure is proposed, enabling saturation-free optimal output regulation while effectively mitigating both modeled and unmodeled disturbances. An active fault-tolerant control (FTC) approach based on the anti-saturation policy algorithm is also introduced to compensate for actuator faults and structural uncertainty. The salient feature of the proposed algorithm is its ability to prevent saturation during optimal control policy learning, while improving both steady-state and transient performance. Finally, simulation studies are conducted to validate the effectiveness of the proposed approach.

This article investigates resilient source seeking problem of second-order multirobot systems (MRSs) under mixed cyberattacks, which consist of misbehaving and Denial-of-Service (DoS) attacks. The misbehaving attacks can cover several types of malicious attacks, such as false data injection, stubborn, and Byzantine, while the network connectivity may be compromised by DoS attacks, potentially resulting in a time-varying and disconnected digraph. To this end, a resilient source seeking algorithm is proposed by designing an auxiliary point for each agent such that the coordination problem is transformed into a point tracking one. A reference velocity is calculated to guide benign robots toward the source, leveraging their historically optimal positions with the highest signal strength. This ensures the auxiliary points converge to the source, clustering benign robots nearby. When DoS attacks occur on some edges, the latest sampling data acquired before the attacks is used to hold the control signals for the robots. Then, sufficient conditions are established through rigorous stability analysis. In comparison to existing methods, the proposed approach extends the safe-kernel-based resilient consensus algorithms to a resilient source seeking algorithm for a general discrete-time second-order dynamics, while also can withstand a mixed cyberattack comprising both misbehaving and DoS attacks. Finally, simulation and experimental results are presented to validate the efficacy of the proposed algorithm.

Formation control of autonomous underwater vehicles (AUVs) has been regarded as the basis of many sophisticated marine missions. However, the complex marine environment and the high communication energy consumption make it hard to achieve this task. This article is concerned with an energy-efficiency formation issue of AUVs via angle measurement. Particularly, the single vector hydrophone is used to measure the single-frequency signal emitted by AUVs, through which the relative angles among AUVs can be estimated. Based on this, a minimally angle rigid topology generation algorithm is designed to balance the tradeoff between communication energy efficiency and topology connectivity, while a model-free inverse reinforcement learning (IRL)-based formation controller is developed to steer AUVs to reach the target while maintaining a specific shape. The innovations are summarized as follows: 1) the angle measurement in this article can eliminate the reliance on the position information of AUVs; 2) the minimally angle rigid topology in this article can improve the formation stability and reduce the communication energy consumption as compared to the neighboring rule-based solutions; and 3) the IRL-based controller can avoid manually designing cost functions and improve environmental adaptability as compared to traditional-learning-based controllers. Finally, simulation and experimental results are both conducted to verify the effectiveness.

The topside separation system is an important device installed on offshore oil exploration platforms for the treatment of produced water. Due to its operation in high-moisture and salt-infested environments, the system is susceptible to valve malfunctions. Additionally, the presence of strong couplings and slugging disturbances in the system further complicate the development of fault-tolerant control (FTC). To achieve this, this article investigates the fault-tolerant$ H_{\infty } $control problem in the topside separation system. To recover control performance against actuator faults while reducing disturbance sensitivity, the fault-tolerant$ H_{\infty } $control problem is formulated for the topside separation system and is expressed as a two-player differential game problem. A Nash equilibrium solution to the fault-tolerant$ H_{\infty } $control problem is derived by solving the game algebraic Riccati equation (GARE). Considering the tailor-made property and difficulty in full-state sensing in industry, an output feedback reinforcement learning (RL) algorithm is proposed to implement the fault-tolerant$ H_{\infty } $control method without the need for system dynamics. Simulation studies are performed to verify the effectiveness of the proposed algorithm.

Pathological images play a vital role in clinical cancer diagnosis. Computer-aided diagnosis utilized on digital Whole Slide Images (WSIs) has been widely studied. The major challenge of using deep learning models for WSI analysis is the huge size of WSI images and existing methods struggle between end-to-end learning and proper modeling of contextual information. Most state-of-the-art methods utilize a two-stage strategy, in which they use a pre-trained model to extract features of small patches cut from a WSI and then input these features into a classification model. These methods can not perform end-to-end learning and consider contextual information at the same time. To solve this problem, we propose a framework that models a WSI as a pathologist's observing video and utilizes Transformer to process video clips with a divide-and-conquer strategy, which helps achieve both context-awareness and end-to-end learning. Extensive experiments on three public WSI datasets show that our proposed method outperforms existing SOTA methods in both WSI classification and positive region detection.

The expensive fine-grained annotation and data scarcity have become the primary obstacles for the widespread adoption of deep learning-based Whole Slide Images (WSI) classification algorithms in clinical practice. Unlike few-shot learning methods in natural images that can leverage the labels of each image, existing few-shot WSI classification methods only utilize a small number of fine-grained labels or weakly supervised slide labels for training in order to avoid expensive fine-grained annotation. They lack sufficient mining of available WSIs, severely limiting WSI classification performance. To address the above issues, we propose a novel and efficient dual-tier few-shot learning paradigm for WSI classification, named FAST. FAST consists of a dual-level annotation strategy and a dual-branch classification framework. Firstly, to avoid expensive fine-grained annotation, we collect a very small number of WSIs at the slide level, and annotate an extremely small number of patches. Then, to fully mining the available WSIs, we use all the patches and available patch labels to build a cache branch, which utilizes the labeled patches to learn the labels of unlabeled patches and through knowledge retrieval for patch classification. In addition to the cache branch, we also construct a prior branch that includes learnable prompt vectors, using the text encoder of visual-language models for patch classification. Finally, we integrate the results from both branches to achieve WSI classification. Extensive experiments on binary and multi-class datasets demonstrate that our proposed method significantly surpasses existing few-shot classification methods and approaches the accuracy of fully supervised methods with only 0.22% annotation costs. All codes and models will be publicly available on https://github.com/fukexue/FAST.

Weakly supervised whole slide image classification is usually formulated as a multiple instance learning (MIL) problem, where each slide is treated as a bag, and the patches cut out of it are treated as instances. Existing methods either train an instance classifier through pseudo-labeling or aggregate instance features into a bag feature through attention mechanisms and then train a bag classifier, where the attention scores can be used for instance-level classification. However, the pseudo instance labels constructed by the former usually contain a lot of noise, and the attention scores constructed by the latter are not accurate enough, both of which affect their performance. In this paper, we propose an instance-level MIL framework based on contrastive learning and prototype learning to effectively accomplish both instance classification and bag classification tasks. To this end, we propose an instance-level weakly supervised contrastive learning algorithm for the first time under the MIL setting to effectively learn instance feature representation. We also propose an accurate pseudo label generation method through prototype learning. We then develop a joint training strategy for weakly supervised contrastive learning, prototype learning, and instance classifier training. Extensive experiments and visualizations on four datasets demonstrate the powerful performance of our method. Codes will be available.

Formation control of remotely operated vehicles (ROVs) has been regarded as the basis of many sophisticated marine missions. However, the high communication energy consumption and weak environment perception ability on ROVs make it challenging to achieve this task. To overcome the above challenge, this article develops a digital twin (DT)-driven formation control approach for ROVs. We first establish a virtual twin model for each ROV by extracting the motion parameters and environment information. With the collected states from ROVs, an integral reinforcement learning (IRL) based formation controller is designed to drive the motion outputs of DT model. After that, the optimal control policy from the DT model is employed to accomplish formation task for each ROV. To reduce the matching error and ensure the formation stability, an IRL-based optimization algorithm is conducted by using the data interaction between DT model and ROVs. Note that the DT-driven formation solution not only can reduce the communication energy consumption by periodically feeding back the real-data of ROVs to the DT model, but also can improve the perception ability of ROVs by reconstructing a virtual twin environment. Finally, experimental results are provided to verify the effectiveness of our solution.

Histopathology image classification is an important clinical task, and current deep learning-based whole-slide image (WSI) classification methods typically cut WSIs into small patches and cast the problem as multi-instance learning. The mainstream approach is to train a bag-level classifier, but their performance on both slide classification and positive patch localization is limited because the instance-level information is not fully explored. In this article, we propose a negative instance-guided, self-distillation framework to directly train an instance-level classifier end-to-end. Instead of depending only on the self-supervised training of the teacher and the student classifiers in a typical self-distillation framework, we input the true negative instances into the student classifier to guide the classifier to better distinguish positive and negative instances. In addition, we propose a prediction bank to constrain the distribution of pseudo instance labels generated by the teacher classifier to prevent the self-distillation from falling into the degeneration of classifying all instances as negative. We conduct extensive experiments and analysis on three publicly available pathological datasets: CAMELYON16, PANDA, and TCGA, as well as an in-house pathological dataset for cervical cancer lymph node metastasis prediction. The results show that our method outperforms existing methods by a large margin. Code will be publicly available.

This article focuses on the safety flocking control problem of networked Lagrangian systems under event-triggered communication. To avoid collision and preserve connectivity, the concept of safety domain and safety vector is presented. Using these definitions, a novel potential function is designed to guarantee the interagent distance to be limited in an exact range during the flocking process. In addition, a fully distributed dynamic event-triggered scheme is proposed to schedule the communication sources. Both the controller and the event-triggered scheme operate without requiring any global information about the entire topology. Also, Zeno behavior is eradicated in the proposed event-triggered communication scheme. Finally, the effectiveness of the proposed method is shown by case studies.

This paper introduces the novel concept of few-shot weakly supervised learning for pathology Whole Slide Image (WSI) classification, denoted as FSWC. A solution is proposed based on prompt learning and the utilization of a large language model, GPT-4. Since a WSI is too large and needs to be divided into patches for processing, WSI classification is commonly approached as a Multiple Instance Learning (MIL) problem. In this context, each WSI is considered a bag, and the obtained patches are treated as instances. The objective of FSWC is to classify both bags and instances with only a limited number of labeled bags. Unlike conventional few-shot learning problems, FSWC poses additional challenges due to its weak bag labels within the MIL framework. Drawing inspiration from the recent achievements of vision-language models (V-L models) in downstream few-shot classification tasks, we propose a two-level prompt learning MIL framework tailored for pathology, incorporating language prior knowledge. Specifically, we leverage CLIP to extract instance features for each patch, and introduce a prompt-guided pooling strategy to aggregate these instance features into a bag feature. Subsequently, we employ a small number of labeled bags to facilitate few-shot prompt learning based on the bag features. Our approach incorporates the utilization of GPT-4 in a question-and-answer mode to obtain language prior knowledge at both the instance and bag levels, which are then integrated into the instance and bag level language prompts. Additionally, a learnable component of the language prompts is trained using the available few-shot labeled data. We conduct extensive experiments on three real WSI datasets encompassing breast cancer, lung cancer, and cervical cancer, demonstrating the notable performance of the proposed method in bag and instance classification. All codes will be made publicly accessible.

The resilient defense strategy for the false data injection attacks (FDIAs) from the perspective of structural vulnerability of smart grids, by using graph and cybernetic approaches, is investigated in this article. FDIA is a kind of well-designed cyberattack that can bypass traditional bad data detection methods and, thus, cause serious damage to smart grids. To defend against FDIAs, a resilient defense control strategy based on virtual hidden networks is proposed to improve the structural vulnerability of smart grids in this study. First, a reduced-dimensional local consensus dynamic model is derived by the Kron reduction method. Second, a virtual hidden network interconnected with the grid is constructed and designed by using the graph theory to indirectly improve the structural vulnerability of the smart grid. In addition, a competitive interconnection approach based on the network zero-sum game is used to further improve the structural vulnerability of the smart grid. The resilience of the smart grid can be effectively improved by changing the topology of the virtual hidden network and the interconnection bipartite graph. Third, the stability of the virtual hidden network-based defense controller is demonstrated separately under either cases without or with FDIAs. The results show that the addition of the virtual hidden network does not change the steady-state operating point of the original grid, but can effectively mitigate the impact of FDIAs on the smart grid. Finally, the effectiveness of the proposed approach is demonstrated on the IEEE 14-bus grid and on the New England 39-bus grid, respectively.

Localization is an indispensable service for underwater sensor networks (USNs). Generally, the convex optimization method is adopted to solve the localization problem. However, the acoustic ray in water medium does not propagate along a straight line, which makes it difficult or impossible to transform the nonconvex optimization problem into a convex optimization problem. This article develops a broad learning (BL)-based localization solution for USNs with isogradient sound speed profile. We first employ the ray tracing model to compensate the range bias caused by straight-line propagation. On the basis of collected range information from anchor nodes, the localization optimization problem is transformed into supervised, unsupervised, and semisupervised learning frameworks. Correspondingly, three BL-based location estimators are developed to seek the position information of sensor nodes, where the incremental learning schemes are conducted for fast parameter tuning and remodeling. In addition, the Cramer–Rao lower bound (CRLB) of positioning error and the convergence to global optimality are both analyzed. Finally, simulation and experiment results are presented to show the effectiveness of our approach. It is demonstrated that the proposed solution in this article has the following nice features: 1) relax the dependence of convex relaxation over convex optimization-based location estimators and 2) reduce the training time and improve the localization efficiency over deep-learning-based location estimators.

3D point cloud registration is a fundamental problem in computer vision and robotics. Recently, learning-based point cloud registration methods have made great progress. However, these methods are sensitive to outliers, which lead to more incorrect correspondences. In this paper, we propose a novel deep graph matching-based framework for point cloud registration. Specifically, we first transform point clouds into graphs and extract deep features for each point. Then, we develop a module based on deep graph matching to calculate a soft correspondence matrix. By using graph matching, not only the local geometry of each point but also its structure and topology in a larger range are considered in establishing correspondences, so that more correct correspondences are found. We train the network with a loss directly defined on the correspondences, and in the test stage the soft correspondences are transformed into hard one-to-one correspondences so that registration can be performed by a correspondence-based solver. Furthermore, we introduce a transformer-based method to generate edges for graph construction, which further improves the quality of the correspondences. Extensive experiments on object-level and scene-level benchmark datasets show that the proposed method achieves state-of-the-art performance.

Learning-based rigid point cloud registration (RPCR) studies have made great progress recently but most existing methods have a small convergence region and can only be used to solve the registration problem with a small rotation angle, which is usually constrained within$[0, 45^\circ ]$. However, the relative rotation between point clouds is usually unconstrained in practice. To address this challenging problem, we propose a new RPCR network and integrate it into a new dual-branch registration framework for unconstrained rotation point cloud registration. The dual-branch framework consists of a large-rotation branch and a small-rotation branch, which are used to accurately register point clouds with large and small relative rotations, respectively. In addition, we propose a multiview intersection over the union module to select a better registration result from the output of the two branches. Extensive experiments on both ModelNet40 and MVP-RG datasets demonstrate that our proposed method outperforms existing state-of-the-art techniques by a large margin.

In this article, the resilient coordination problem of networked Lagrangian systems with adversarial agents is considered. A novel algorithm called norm-based resilient decision algorithm is proposed to exclude the impact of adversarial agents. To ensure the coordination of networked Lagrangian systems, the maximum number of adversarial agents related to the robustness of the communication network is given. Under the proposed resilient consensus algorithm, secure coordination is guaranteed under adversarial agents. Then, the proposed resilient controller is extended to static formation scenarios. Finally, the effectiveness of the proposed method is demonstrated through case studies and experiments. Compared to the existing results, the proposed algorithm can reduce computing resources by designing auxiliary vectors and converting them into scalars to remove extreme values.

Communication efficiency and flocking stability are two basic requirements for the application of autonomous underwater vehicles (AUVs) in maritime transportation systems. Although they are closely related, most existing flocking approaches focus on the control techniques and ignore the influence of communication efficiency. This paper presents a joint design solution to the channel estimation and flocking control for multi-AUV-based maritime transportation systems, with the consideration of path loss, shadow and multipath fading channels. A value iteration-based reinforcement learning (RL) estimator is first designed to predict the channel quality of AUVs in positions that have not yet visited. With the predicted channel quality, we construct an integrated optimization problem for the co-design of communication and flocking strategies. Along with this, a value iteration-based RL flocking controller is developed to achieve the co-design of channel estimation and flocking control for AUVs. It is worth mentioning that, the value iteration-based RL estimator in this paper can avoid local optimal in traditional least-squares methods, and meanwhile the flocking controller in this paper can make a balance between flocking stability and communication efficiency for AUVs. Finally, simulation and experimental results reveal that the proposed approach in this paper has superior performances by comparing with the other works. As such, our approach is more useful for marine engineer to understand and explore the maritime transportation system from the communication and control view points.

This article is concerned with a containment control issue for autonomous underwater vehicles (AUVs), subject to unavailable velocity signals in cyber side and stochastic environment disturbances in physical side. We first divide the environmental disturbances into deterministic and stochastic parts. Based on this, a terminal sliding mode observer is developed to estimate the velocities of AUVs in finite time. With the estimated velocities, a distributed containment controller is designed for each AUV to follow a convex hull spanned by trajectories of the leader AUVs. For the developed velocity observer, a double power reaching law is employed to reduce the chattering and improve the convergence rate. Besides that, an adaptive strategy is incorporated into the containment controller, such that the steady-state errors caused by stochastic environment disturbances can be compensated. Stability conditions for the velocity observer and containment controller are also provided. Finally, we conduct the simulation and experimental studies to verify the effectiveness.

Fast and smooth driving is a preferable consideration in platooning algorithm development for intelligent autonomous vehicular systems. It can improve traffic efficiency while guaranteeing passenger comfort. To this end, this work investigates the fast distributed platooning problem of connected vehicular systems. Taking the inaccurate velocity measurement into consideration, an extended state observer (ESO) based on a fractional order faster nonsingular terminal sliding mode (FNTSM) is proposed for velocity and disturbance estimation simultaneously. An FNTSM control algorithm based on the double power reaching law is developed to reach fast platooning while guaranteeing string stability of the connected vehicular systems regardless of zero or nonzero initial spacing error conditions. The salient features of the proposed platoon controller are that system convergence can be achieved in finite time and the time-varying external disturbances can be estimated accurately. Finally, simulation and experiment studies are conducted to demonstrate the effectiveness and efficiency of the proposed control algorithm.

Computer-aided pathology diagnosis based on the classification of Whole Slide Image (WSI) plays an important role in clinical practice, and it is often formulated as a weakly-supervised Multiple Instance Learning (MIL) problem. Existing methods solve this problem from either a bag classification or an instance classification perspective. In this paper, we propose an end-to-end weakly supervised knowledge distillation framework (WENO) for WSI classification, which integrates a bag classifier and an instance classifier in a knowledge distillation framework to mutually improve the performance of both classifiers. Specifically, an attention-based bag classifier is used as the teacher network, which is trained with weak bag labels, and an instance classifier is used as the student network, which is trained using the normalized attention scores obtained from the teacher network as soft pseudo labels for the instances in positive bags. An instance feature extractor is shared between the teacher and the student to further enhance the knowledge exchange between them. In addition, we propose a hard positive instance mining strategy based on the output of the student network to force the teacher network to keep mining hard positive instances. WENO is a plug-and-play framework that can be easily applied to any existing attention-based bag classification methods. Extensive experiments on five datasets demonstrate the efficiency of WENO. Code is available at https://github.com/miccaiif/WENO.

In this article, the predefined-time cooperative platoon control problem with constrained communication range for uncertain networked aerial vehicles is considered. The uncertain networked aerial vehicles are subjected to external disturbance and parameter uncertainties. A novel adaptive sliding-mode disturbance observer is first designed for single uncertain networked aerial vehicle with an radial basis function of neural network estimator to guarantee the control performance. By embedding a dynamic control gain associated with the predefined convergence time, the proposed disturbance observer is proved to be uniformly ultimate boundedness stable by the time transformation approach. Then the dynamic gain technology is combined with the prescribed performance control to design the platoon controllers for uncertain networked aerial vehicles. With the proposed disturbance observer and the distribute controller, the platoon can be achieved within the predefined time. The proposed approach can simultaneously guarantee the system stabile without any initial conditions and system parameters. Finally, some simulation experiments are given to verify the effectiveness of the proposed protocols.

In this article, a coupled sliding mode control (CSMC) is developed for vehicular systems with nonlinear uncertainties by using the disturbance observer (DO) and multi power reaching law. The DO is designed to estimate the nonlinear uncertainties. It is worth mentioning for the DO that the uncertainties include not only parameter uncertainty but also external disturbance, and the bounds of the uncertainties are not required to be known. In addition, the multi power reaching law is constructed to avoid the chattering problem of the traditional sliding mode control (SMC) and to improve the convergence speed effectively. Firstly, the constant time headway policy (CTHP) based on multi power reaching law and SMC is proposed to achieve the string stability for vehicle platoons. Compared with constant spacing policy (CSP), CTHP is more feasible in practice, because the desired spacing between adjacent vehicles is dependent on vehicle speed. Then, a modified constant time headway policy (MCTHP) is proposed for the vehicular systems to decrease the intervehicle spacing and increase the traffic density effectively. Finally, the numerical simulation and experiment are performed to demonstrate the effectiveness and advantage of the developed strategy.

Human-on-the-loop (HOTL) system is regarded as a promising technology to allow autonomous underwater vehicle (AUV) to track the most adequate target point as soon as possible. However, the unique characteristics of the underwater environment make it challenging to perform the tracking task. This article is concerned with a finite-time tracking control issue for AUV, subjected to unavailable velocity signals in the measurement side and uncertain model parameters in physical side. A HOTL system, including operator, buoys, AUV and sensors, is first provided to construct a cooperative tracking network. For such system, operator in surface control center decides the tracking mission based on all available data. Then, a buoy-assisted localization estimator is utilized by AUV to acquire its position, through which a fast terminal sliding mode observer is developed to estimate the velocity of AUV in finite time. With the estimated velocity information, an adaptive-nonsingular fast terminal sliding mode tracking controller is designed to drive AUV to the target point in finite time. For the proposed velocity observer and tracking controller, the signum and differential functions are employed together to improve the convergence speed and reduce the chattering. Besides that, the proposed solution can not only guarantee finite-time velocity observation, but also achieve finite-time tracking control. Finally, simulation and experimental results are both presented to verify the effectiveness.

This article studies the joint localization and tracking issue for the autonomous underwater vehicle (AUV), with the constraints of asynchronous time clock in cyberchannels and model uncertainty in physical channels. More specifically, we develop a reinforcement learning (RL)-based asynchronous localization algorithm to localize the position of AUV, where the time clock of AUV is not required to be well synchronized with the real time. Based on the estimated position, a scalable sampling strategy called multivariate probabilistic collocation method with orthogonal fractional factorial design (M-PCM-OFFD) is employed to evaluate the time-varying uncertain model parameters of AUV. After that, an RL-based tracking controller is designed to drive AUV to the desired target point. Besides that, the performance analyses for the integration solution are also presented. Of note, the advantages of our solution are highlighted as: 1) the RL-based localization algorithm can avoid local optimal in traditional least-square methods; 2) the M-PCM-OFFD-based sampling strategy can address the model uncertainty and reduce the computational cost; and 3) the integration design of localization and tracking can reduce the communication energy consumption. Finally, simulation and experiment demonstrate that the proposed localization algorithm can effectively eliminate the impact of asynchronous clock, and more importantly, the integration of M-PCM-OFFD in the RL-based tracking controller can find accurate optimization solutions with limited computational costs.

3D point cloud registration is a fundamental problem in computer vision and robotics. Recently, learning-based point cloud registration methods have made great progress. However, these methods are sensitive to outliers, which lead to more incorrect correspondences. In this paper, we propose a novel deep graph matching-based framework for point cloud registration. Specifically, we first transform point clouds into graphs and extract deep features for each point. Then, we develop a module based on deep graph matching to calculate a soft correspondence matrix. By using graph matching, not only the local geometry of each point but also its structure and topology in a larger range are considered in establishing correspondences, so that more correct correspondences are found. We train the network with a loss directly defined on the correspondences, and in the test stage the soft correspondences are transformed into hard one-to-one correspondences so that registration can be performed by singular value decomposition. Furthermore, we introduce a transformer-based method to generate edges for graph construction, which further improves the quality of the correspondences. Extensive experiments on registering clean, noisy, partial-to-partial and unseen category point clouds show that the proposed method achieves state-of-the-art performance. The code will be made publicly available at https://github.com/fukexue/RGM.

The cyber security of large-scale smart grid against false data injection attack (FDIA) is concerned in this article. FDIA can modify the sensor data and make internal states cause bias without being detected by the bad data detection system. We propose a method for FDIA detection and localization in the smart grid in this article. First, a series of interval observers are designed by considering the bounds of internal states, modeling errors, and disturbances to estimate the interval states of the grid physical system. By using the interval residuals of interval observers, a detection scheme against FDIA is proposed. For FDIA localization, the measurement data of the corresponding sensor is used as the input of the interval observer. Therefore, each interval observer is responsible for FDIA detection and localization of the corresponding sensor. Furthermore, the logic localization judgment matrix is constructed for localizing the sensor in which FDIA is injected. Then, the detection and localization scheme against FDIA is proposed based on the interval observer and the logic localization judgment matrix. Finally, simulations on the IEEE 36-bus grid are performed to illustrate the effectiveness of the proposed interval observer-based FDIA detection and localization algorithm.

Privacy-preserving localization for Internet of Underwater Things (IoUT) plays a fundamental role in the sensing, communication and control of ocean environments. However, the unique characteristics of underwater environment make it much more difficult to achieve such a task. In this article, we are concerned with a privacy-preserving localization issue for IoUT, subjected to asynchronous clock, stratification effect and forging attack in cyber channels. In order to eliminate the influence of asynchronous clock and hide the private position information, we develop a privacy-preserving asynchronous transmission protocol, where a received signal strength (RSS)-based detection strategy is given to detect the malicious anchor nodes. Based on this, a least squares estimator is designed to estimate the position information of target. Particularly, a ray compensation strategy is incorporated into the localization estimator, such that the localization bias from assuming the straight-line transmission can be avoided. It is worth mentioning that, the proposed localization solution in this article can not only hide the private position information, but also eliminate the influences of asynchronous clock, stratification effect and forging attack. Finally, simulation and experiment results are conducted to reveal that the proposed localization solution outperforms the other existing works in terms of localization accuracy and effectiveness.

Tracking an autonomous underwater vehicle (AUV) has been regarded as one of the most key applications for Internet of Underwater Things (IoUT). However, the strong mobility of AUV as well as asynchronous clock, stratification effect, and high energy consumption of acoustic communication make it challenging to achieve such a task. To handle the above issues, this article develops a ubiquitous tracking scheme for AUV. The tracking scheme is divided into two stages, i.e.: 1) motion prediction and 2) persistent tracking. In the first stage, an unscented transform-based localization estimator is utilized by sensor nodes to acquire the initial position of AUV, through which a terminal sliding-mode velocity observer is designed to predict the mobility trajectory of AUV. With the predicted mobility trajectory, a minimum rigid-graph-based tracking strategy is developed in the second stage to enable ubiquitously tracking. For the designed tracking strategy, the posterior Cramer-Rao lower bound is selected as the benchmark to optimize the network topology, such that a minimum rigid graph can be generated to balance the tradeoff between tracking accuracy and energy consumption. Particularly, the duty-cycle mechanism and the unscented Kalman filtering are jointly adopted to prolong the network lifetime and improve the tracking accuracy. Finally, simulation and experimental results are presented to show the effectiveness of our approach.

Underwater sensor networks (USNs) are envisioned to enable a large variety of marine applications. Such applications require accurate position information of sensor nodes. However, the openness and inhomogeneity characteristics of underwater medium make it much more challenging to solve the localization issue. This paper is concerned with a privacy-preserving localization issue for USNs in inhomogeneous underwater medium. An honest-but-curious model is considered to develop a privacy-preserving localization protocol. Based on this, a localization problem is constructed for sensor nodes to minimize the sum of all measurement errors, where a ray compensation strategy is incorporated to remove the localization bias from assuming the straight-line transmission. To make the above problem tractable, we consider the unsupervised, supervised and semisupervised scenarios, through which deep reinforcement learning (DRL) based localization estimators are utilized to estimate the positions of sensor nodes. It is noted that, the proposed localization solution in this paper can hide the private position information of USNs, and more importantly, it is robust to local optimum for nonconvex and nonsmooth localization problem in inhomogeneous underwater medium. Finally, simulation studies are given to show the position privacy can be preserved, while the localization accuracy can be enhanced as compared with the other existing works.

This article investigates the detection and isolation of false data injection (FDI) attacks in a smart grid based on the unknown input (UI) interval observer. Recent studies have shown that the FDI attacks can bypass the traditional bad data detection methods by using the vulnerability of state estimation. For this reason, the emergency of FDI attacks brings enormous risk to the security of smart grid. To solve this crucial problem, an UI interval observer-based detection and the isolation scheme against FDI attacks are proposed. We first design the UI interval observers to obtain interval state estimation accurately, based on the constructed physical dynamics grid model. Through the capabilities of the designed UI interval observers, the accurate interval estimation state can be decoupled from unknown disturbances. Based on the characteristics of the interval residuals, a UI interval observer-based global detection algorithm was proposed. Particularly, the interval residual-based detection criteria can address the limitation of the precomputed threshold in traditional bad data detection methods. On this basis, we further consider the detection and isolation of FDI attacks under structure vulnerability. Namely, there exist undetectable FDI attacks in the grid system. Taking the attack undetectability problem into account, a logic judgment matrix-based local detection and isolation algorithm against FDI attacks are developed. Based on the combinations of observable sensor cases, local control centers can further detect and isolate the attack set under structure vulnerability. Finally, the effectiveness of the developed detection and isolation algorithms against FDI attacks is demonstrated on the IEEE 8-bus and IEEE 118-bus smart grid system, respectively.

Localization is a critical issue for many location-based applications in the Internet of Underwater Things (IoUT). Nevertheless, the asynchronous time clock, stratification effect, and mobility properties of the underwater environment make it much more challenging to solve the localization issue. This article is concerned with an autonomous underwater vehicle (AUV)-aided localization issue for IoUT. We first provide a hybrid network architecture that includes surface buoys, AUVs, and active and passive sensor nodes. On the basis of this architecture, an asynchronous localization protocol is designed, through which the localization problem is provided to minimize the sum of all measurement errors. In order to make this problem tractable, a reinforcement-learning (RL)-based localization algorithm is developed to estimate the locations of AUVs, and active and passive sensor nodes, where an online value iteration procedure is performed to seek the optimization locations. It is worth mentioning that the proposed localization algorithm adopts two neural networks to approximate the increment policy and value function, and more importantly, it is much preferable for the nonsmooth and nonconvex underwater localization problem due to its insensitivity to the local optimal. Performance analyses for the RL-based localization algorithm are also provided. Finally, simulation and experimental results reveal that the localization performance in this article can be significantly improved as compared with the other works.

Target tracking has been considered as one of the most important applications of underwater acoustic sensor networks. However, the long propagation delay, high-energy consumption, and strong noise properties of the underwater environment make target tracking more challenging as compared with terrestrial sensor networks. This article is concerned with an energy-efficient tracking issue for underwater targets, subject to an asynchronous clock, power restriction, and noise measurement constraints. The tracking process can be divided into two phases, i.e., position acquisition and persistent tracking. In the first phase, we establish the relationship between propagation delay and position, through which an asynchronous localization algorithm is developed for sensor nodes to estimate the position of target. Based on the estimated position, a consensus-based Bayesian filter is designed for sensor nodes in the second phase to enable persistent tracking. In particular, the consensus fusion strategy and duty-cycle mechanism are jointly adopted to improve the tracking accuracy and prolong the network lifetime. Moreover, the convergence analyses for the proposed approach are also presented. Finally, simulation and experimental results reveal that the proposed tracking approach can reduce the influence of malicious measurements, while the energy efficiency can be significantly improved as compared with the other works.

The detection and isolation problem of false data injection (FDI) attacks in large-scale smart grid systems, is investigated in this paper. The FDI attacks can bypass the traditional bad data detection techniques, by falsifying the process of state estimation. For this reason, the emergency of FDI attacks brings great risk to the security of smart grids. To address this crucial problem, a novel detection and isolation scheme against the FDI attacks for the large-scale smart grid system is proposed. We first design an interval observer to estimate the interval state of internally physical system accurately, based on the constructed physical dynamics of grid systems. Taking the bounds of internal state and external disturbance into account, the detection criterion that an alarm is generated when the interval residuals does not include the zero value is proposed. To address the limitation of precomputed threshold, we use the interval residuals regarded as a nature detection threshold to replace the evaluation function and detection threshold used in traditional attack detection methods. Furthermore, an attack signature logical judgment matrix-based isolation algorithm is further proposed to isolate the sensors, in which the FDI attacks may be injected into the attacked subarea. Finally, the effectiveness of the developed detection and isolation scheme is demonstrated by using detailed case studies on the IEEE 128-bus smart grid system.

Directional antenna-based aerial networking (DAAN) is referred as a promising technology to meet the dynamic data demands for unmanned aerial vehicles (UAVs). However, the narrow radiation pattern of directional antennas and the mobility of UAVs make it challenging to form a robust DAAN. This paper presents a heading control strategy for UAV-carried directional antennas to establish a robust long-range aerial communication channel. The heading control process is mainly divided into two phases, i.e., position estimation and angle adjustment. In the first phase, a proportional-derivative-based tracking controller is designed for each UAV to ensure the consistency of heights, pitch, and roll angles. Particularly, the received signal strength indicator is adopted as an auxiliary measuring component, and then a consensus-based unscented Kalman filtering algorithm is developed to estimate the position of UAVs. With the estimated position information, a feedback-based heading controller is designed for directional antenna in the second phase to enable robust long-range aerial communication channel. Moreover, the convergence conditions and Cramér-Rao lower bounds are also provided. Finally, simulation results are presented to demonstrate the effectiveness of the proposed strategy. It is shown that the influence of malicious measurements can be reduced, and the signal strength can be significantly improved as compared with the omni-directional antenna-based works.

This letter is concerned with an asynchronous localization issue for underwater acoustic sensor networks (UASNs), subject to asynchronous clocks and stratification effects in physical channels. A novel unscented transform-based localization algorithm is proposed to estimate the positions of sensor nodes. Instead of linearizing the measurement equations, the proposed algorithm employs the unscented transform to compute the Jacobian matrix to reduce the linearization errors. Particularly, the ray-tracing approach is adopted to model the stratification effect. Moreover, the convergence analysis and Cramér-Rao lower bound for the algorithm are also provided. Simulation results show that the proposed algorithm can effectively improve the estimation accuracy as compared with the existing works.

In this paper, the output containment control problem of heterogeneous linear multi-agent systems is investigated. The objective of the output containment control problem is to make a group of agents converge to a convex hull spanned by some leaders. A distributed control law based on output regulation framework is proposed, where a distributed observer is designed in the control law for agents estimating the leaders' states. The salient feature of the proposed control law is that convergence trajectories of the distributed observers can only be determined by system communication topology and the initial states of leaders, which is helpful to obtain local necessary and sufficient condition for the solvability of the output containment control problem. Simulation studies demonstrate that the proposed control law is effective and efficient.

This paper investigates the output consensus problem of heterogeneous discrete-time multiagent systems with individual agents subject to structural uncertainties and different disturbances. A novel distributed control law based on internal reference models is first presented for output consensus of heterogeneous discrete-time multiagent systems without structural uncertainties, where internal reference models embedded in controllers are designed with the objective of reducing communication costs. Then based on the distributed internal reference models and the well-known internal model principle, a distributed control law is further presented for output consensus of heterogeneous discrete-time multiagent systems with structural uncertainties. It is shown in both cases that the consensus trajectory of the internal reference models determines the output trajectories of agents. Finally, numerical simulation results are provided to illustrate the effectiveness of the proposed control schemes.

This paper presents a new adaptive bandwidth allocation scheme to prevent handoff failure in wireless cellular networks, known as the measurement-based preassignment (MPr) technique. This technique is particularly useful in micro/pico cellular networks which offers quality-of-service (QoS) guarantee against call dropping. The proposed MPr scheme distinguishes itself from the well-known guarded channel (GC) based schemes in that it allows the handoff calls to utilize a prereserved channel pool before competing for the shared channels with new call arrivals. The key advantage of the proposed MPr scheme is that it enables easy derivation of the number of channels that needs to be reserved for handoff based on a predetermined handoff dropping probability, without the need for solving the often complex Markov chain required in GC schemes, thus, making the proposed MPr scheme simple and efficient for implementation. This is essential in handling multiple traffic types with potentially different QoS requirements. In addition, the MPr scheme is adaptive in that it can dynamically adjust the number of reserved channels for the handoff according to the periodical measurement of the traffic status within a local cell, thus completely eliminating the signaling overhead for status information exchange among cells mandated in most existing channel allocation schemes. Numerical results and comparisons are given to illustrate the tradeoff.

This paper presents a new dynamic and adaptive bandwidth allocation scheme to prevent handoff failure due to lack of resources in mobile cellular networks, known as the measurement-based preassignment (MPr) technique. This technique is particularly useful in micro/pico cellular networks which offer quality of service (QoS) guarantees against call dropping. The proposed MPr scheme distinguishes itself from the well-known guarded channel based schemes in that it allows the handoff calls to utilize a pre-reserved channel pool before competing for the shared channels with new call arrivals. The key advantage of the proposed MPr scheme is that it can easily derive the number of channels that needs to be reserved for handoff based on a pre-determined call dropping probability, without the need for solving the often complex Markov chain required in guarded channel schemes, thus making the proposed MPr scheme simple and efficient for implementation. In addition, the MPr scheme is dynamical in that it can adjust the number of reserved channels for the handoff according to the traffic changes in the system.

It is foreseeable that a micro/pico-cellular architecture will be the most attractive wireless communication solution in the future because of the high frequency reuse the structure can support. Due to smaller cell size, the rate of call handoffs should increase. From a user's perspective, having a call abruptly terminated in the middle of a conversation is far more annoying than having a new call attempt blocked. In this paper a dynamic channel pre-reservation (DCPr) scheme based on incoming handoff rate and local carried load is proposed as a means of providing handoff priority. The proposed DCPr scheme is capable of providing grade of service (GoS) guarantee in terms of call dropping probability. In addition, operation for base stations is very simple, which makes the proposed scheme practical to implement in mobile micro/pico-cellular networks.