This paper is concerned with the distributed state estimation problem over wireless sensor networks. The communication links are unreliable that are subject to random link failures modeled as a set of independent Bernoulli processes. To estimate the plant state collaboratively, a Kalman-consensus filtering approach is developed where the sensors spread the local information obtained from the Kalman filtering algorithm by performing a consensus of the inverse covariance matrices at each time instant. Sufficient conditions for the stochastic boundedness of the Kalman-consensus filter are established. It is shown that the filtering performance is directly influenced by the network connectivity and the collective observability. A numerical example is illustrated to verify the proposed results.

This article is concerned with moving-horizon state estimation problems for a class of discrete-time linear dynamic networks. The signals are transmitted via noisy network channels and distortions can be caused by channel noises. As such, the binary encoding schemes, which take advantages of the robustness of the binary data, are exploited during the signal transmission. More specifically, under such schemes, the original signals are encoded into a bit string, transmitted via memoryless binary symmetric channels with certain crossover probabilities, and eventually restored by a decoder at the receiver. Novel centralized and decentralized moving-horizon estimators in the presence of the binary encoding schemes are constructed by solving the respective global and local least-square optimization problems. Sufficient conditions are obtained through intensive stochastic analysis to guarantee the stochastically ultimate boundedness of the estimation errors. A simulation example is presented to verify the effectiveness of the proposed moving-horizon estimators.

In this paper, we consider the distributed filtering problem for continuous-time stochastic systems over sensor networks subject to Markovian switching topologies. Due to limited communication energy and bandwidth, an event-based communication scheme is proposed with the aim to decrease the transmission frequency. An individual triggering condition is put forward to regulate the communication rates for each component of the system state in order to better reflect the engineering requirements. The aim of this paper is to design a distributed filter over sensor networks with Markovian switching topologies such that the dynamics of the estimation error is exponentially mean-square bounded. It is shown that, with the proposed event-based distributed filtering algorithm, the exponential mean-square boundedness of the estimation errors is guaranteed if the sensor network is distributively detectable and the combined communication topology is strongly connected. A numerical example is presented to illustrate the usefulness of the developed algorithm.

In this paper, we investigate the state estimation problem for a class of networked systems with energy harvesting technologies, where the sensor is capable of replenishing energy from the environment. The underlying system is subject to both additive and multiplicative stochastic noises, and the measurement is transmitted to the remote estimator only when the current energy storage is larger than the transmission energy consumption. A binary encoding scheme is utilized in the communication process, under which the measurements are quantized into a bit string, transmitted via memoryless binary symmetric channels with certain probabilistic bit flips, and recovered at the receiver. A minmax robust estimator is designed to minimize the worst-case covariance of the estimation error in terms of the solutions to Riccati-like difference equations. Furthermore, the influence of the length of bit stream on the transmission rate and the estimation performance is discussed, and conditions guaranteeing the boundedness of the proposed estimator are provided. Finally, a numerical example is exploited to verify the effectiveness of the results.

In this article, the diagnosability problem is investigated for open-loop and closed-loop stochastic systems under sensor and actuator faults. A novel quantitative fault diagnosability analysis method is developed to characterize the detectability and isolability for a given system without any predetermined fault diagnosis algorithm. By exploiting the Bhattacharyya distance, some novel indices are proposed to evaluate the difficulty degree of fault detectability and isolability. Necessary and sufficient conditions are established for the diagnosability of both open- and closed-loop stochastic systems undergoing sensor and actuator faults. In addition, the relationship between the fault detectability of open-loop systems and that of closed-loop ones is discussed.

This article is concerned with domain generalization (DG), a practical yet challenging scenario in transfer learning where the target data are not available in advance. The key insight of DG is focused on learning a robust model that can generalize to the unseen domain by leveraging knowledge from the source domain. To this end, we propose a novel algorithm known as support-sample-assisted Adversarial Attacks (SSAA) for DG. In the SSAA algorithm, an attack–defense strategy is deployed to enhance the target model's generalizability and transferability. This strategy includes a nontargeted attack stage, during which attack samples are generated to form pseudotarget domains with near-realistic covariate shifts. Subsequently, in the model defense stage, a biclassifier structure is used to distinguish support samples from the generated attack samples. These support samples form a new decision boundary encompassing all unseen samples, prompting an extension of the existing decision boundary to meet these samples. Experimental results on cross-domain fault diagnosis tasks suggest that SSAA outperforms current state-of-the-art DG methods, indicating a promising avenue for further DG development.