This paper investigates the fuzzy distributed resilient formation control problem of Takagi–Sugeno (T–S) fuzzy MASs under DoS attacks and communication delays. Since the communication network among MASs is subject to DoS attacks and communication delays, the leader’s states cannot be obtained continuously by each agent. A fuzzy distributed resilient estimator is first proposed to estimate the leader’s state subject to DoS attacks and communication delays. By constructing a Lyapunov-Krasovskii functional, it is demonstrated that the estimation errors of the fuzzy resilient estimator exponentially converge to zero. Secondly, based on the formulated fuzzy distributed resilient estimator and parallel distributed compensation (PDC) algorithm, a fuzzy distributed resilient formation controller is developed. The presented fuzzy distributed resilient formation control strategy ensures that the controlled T–S fuzzy MASs remain stable and achieve the formation objective. Finally, the developed fuzzy distributed resilient formation control method is applied to the single-link robot arm systems, the effectiveness is demonstrated by simulation and comparison results.

In this article, we study the reinforcement learning (RL) optimal output feedback control problem for Takagi–Sugeno (T–S) fuzzy systems with immeasurable states and disturbances. A fuzzy filtering observer is designed to estimate the immeasurable states, and then, based on the filtering observer, a fuzzy optimal output feedback control method is presented by employing zero-sum differential game theory. Since the analytical optimal control solutions are reduced to solving game algebraic Riccati equations (GAREs), which is difficult to obtain their analytical solutions, an output feedback model-free policy iteration (PI) learning algorithm is proposed. It is proved that the proposed algorithm is convergent and the proposed fuzzy RL optimal output feedback control approach can make the controlled systems be asymptotically stable and satisfy the disturbance attenuation condition. Finally, we apply the developed optimal control method to a mass–spring–damper system, and the simulation results verify the effectiveness of the developed method.

Inverse reinforcement learning optimal control is under the framework of learner-expert, the learner system can learn expert system's trajectory and optimal control policy via a reinforcement learning algorithm and does not need the predefined cost function, so it can solve optimal control problem effectively. This paper develops a fuzzy inverse reinforcement learning optimal control scheme with inverse reinforcement learning algorithm for Takagi-Sugeno (T-S) fuzzy systems with disturbances. Since the controlled fuzzy systems (learner systems) desire to learn or imitate expert system's behavior trajectories, a learner-expert structure is established, where the learner only know the expert system's optimal control policy. To reconstruct expert system's cost function, we develop a model-free inverse Q-learning algorithm that consists of two learning stages: an inner Q-learning iteration loop and an outer inverse optimal iteration loop. The inner loop aims to find fuzzy optimal control policy and the worst-case disturbance input via learner system's cost function by employing zero-sum differential game theory. The outer one is to update learner system's state-penalty weight via only observing expert systems' optimal control policy. The model-free algorithm does not require that the controlled system dynamics are known. It is proved that the designed algorithm is convergent and also the developed inverse reinforcement learning optimal control policy can ensure T-S fuzzy learner system to obtain Nash equilibrium solution. Finally, we apply the presented fuzzy inverse Q-learning optimal control method to nonlinear unmanned surface vehicle system and the computer simulation results verified the effectiveness of the developed scheme.

In this article, a fuzzy optimal event-triggered dynamic positioning control approach with aQ-learning value iteration (VI) algorithm is developed for unmanned surface vehicles (USVs) systems. The USV systems are first modeled by Takagi-Sugeno (T-S) fuzzy systems. To reduce the communication resources and controller update times, an event-triggered mechanism is designed via employing the sampled augmented systems states and triggered control input signals. Based on the developed event-triggered mechanism and Bellman optimality theory, a fuzzy optimal event-triggered control (ETC) approach is presented. Since solution of optimal control policy reduces to algebraic Riccati equations (AREs), its analytical solution is difficult to solve directly. Then, to search its approximation solution, a VI algorithm is formulated. By rigorous proof, the proposed optimal ETC scheme can assure that the USVs systems are asymptotically stable and theQ-learning algorithm is convergent. Finally, the simulation and comparisons results with previous optimal controllers verify the feasibility of the presented optimal ETC scheme.

In this work, we share our experience on tele-knowledge pre-training for fault analysis, a crucial task in telecommunication applications that requires a wide range of knowledge normally found in both machine log data and product documents. To organize this knowledge from experts uniformly, we propose to create a Tele-KG (tele-knowledge graph). Using this valuable data, we further propose a tele-domain language pre-training model TeleBERT and its knowledge-enhanced version, a tele-knowledge re-training model KTeleBERT. which includes effective prompt hints, adaptive numerical data encoding, and two knowledge injection paradigms. Concretely, our proposal includes two stages: first, pre-training TeleBERT on 20 million tele-related corpora, and then re-training it on 1 million causal and machine-related corpora to obtain KTeleBERT. Our evaluation on multiple tasks related to fault analysis in tele-applications, including root-cause analysis, event association prediction, and fault chain tracing, shows that pretraining a language model with tele-domain data is beneficial for downstream tasks. Moreover, the KTeleBERT re-training further improves the performance of task models, highlighting the effectiveness of incorporating diverse tele-knowledge into the model.

Multi-modal entity alignment (MMEA) aims to discover identical entities across different knowledge graphs (KGs) whose entities are associated with relevant images. However, current MMEA algorithms rely on KG-level modality fusion strategies for multi-modal entity representation, which ignores the variations of modality preferences of different entities, thus compromising robustness against noise in modalities such as blurry images and relations. This paper introduces MEAformer, a mlti-modal entity alignment transformer approach for meta modality hybrid, which dynamically predicts the mutual correlation coefficients among modalities for more fine-grained entity-level modality fusion and alignment. Experimental results demonstrate that our model not only achieves SOTA performance in multiple training scenarios, including supervised, unsupervised, iterative, and low-resource settings, but also has a limited number of parameters, efficient runtime, and interpretability. Our code is available at https://github.com/zjukg/MEAformer.

Knowledge graphs (KG) are essential background knowledge providers in many tasks. When designing models for KG-related tasks, one of the key tasks is to devise the Knowledge Representation and Fusion (KRF) module that learns the representation of elements from KGs and fuses them with task representations. While due to the difference of KGs and perspectives to be considered during fusion across tasks, duplicate and ad hoc KRF modules design are conducted among tasks. In this paper, we propose a novel knowledge graph pretraining model KGTransformer that could serve as a uniform KRF module in diverse KG-related tasks. We pretrain KGTransformer with three self-supervised tasks with sampled sub-graphs as input. For utilization, we propose a general prompt-tuning mechanism regarding task data as a triple prompt to allow flexible interactions between task KGs and task data. We evaluate pretrained KGTransformer on three tasks, triple classification, zero-shot image classification, and question answering. KGTransformer consistently achieves better results than specifically designed task models. Through experiments, we justify that the pretrained KGTransformer could be used off the shelf as a general and effective KRF module across KG-related tasks. The code and datasets are available at https://github.com/zjukg/KGTransformer.

The finite-time dynamic event-triggered fuzzy output feedback fault-tolerant control problem is studied in this article for the interval type-2 (IT2) Takagi–Sugeno fuzzy system with parameter uncertainties and actuator faults. A fuzzy state observer is first developed to solve the immeasurable state problem. Second, by using the sampled estimating states and measured output signals, a dynamic event-triggered mechanism is formulated via integrating sensor-to-observer with observer-to-controller. Third, an observer-based finite-time event-triggered fuzzy fault-tolerant controller is synthesized via the nonparallel distribution compensation design principle. Consequently, the finite-time stable conditions of the addressed IT2 fuzzy system are established by constructing an appropriate Lyapunov function. Furthermore, an output feedback control design algorithm of solving control and observer gains is given in terms of the established sufficient finite-time stable conditions. Finally, a practical example of a nonlinear tunnel diode circuit system is provided to verify the effectiveness of the proposed IT2 fuzzy control scheme.

Audio has been increasingly adopted as a sensing modality in a variety of human-centered mobile applications and in smart assistants in the home. Although acoustic features can capture complex semantic information about human activities and context, continuous audio recording often poses significant privacy concerns. An intuitive way to reduce privacy concerns is to degrade audio quality such that speech and other relevant acoustic markers become unintelligible, but this often comes at the cost of activity recognition performance. In this paper, we employ a mixed-methods approach to characterize this balance. We first conduct an online survey with 266 participants to capture their perception of privacy qualitatively and quantitatively with degraded audio. Given our findings that privacy concerns can be significantly reduced at high levels of audio degradation, we then investigate how intentional degradation of audio frames can affect the recognition results of the target classes while maintaining effective privacy mitigation. Our results indicate that degradation of audio frames can leave minimal effects for audio recognition using frame-level features. Furthermore, degradation of audio frames can hurt the performance to some extend for audio recognition using segment-level features, though the usage of such features may still yield superior recognition performance. Given the different requirements on privacy mitigation and recognition performance for different sensing purposes, such trade-offs need to be balanced in actual implementations.

This paper studies the problem of high-speed train operation with special attention to minimizing the energy consumption. The performance characteristics of a high-speed train, including traction characteristic and regenerative braking, and the railway geographical conditions consisting of slope, curve, and tunnel parameters, are fully considered in the dynamic model in order to make it more effective and practical. A new optimal strategy for train operation is developed, and its novelty lies in the fact that it is the first time to optimize the actual speed curve using the method of dual speed curve optimization, which contains two processes of offline global optimization and online local optimization, thus leading to more energy saving as compared with most existing methods with only one-time optimization process. We utilize combination optimization techniques, in tandem with the speed codes and subsections, to solve the global optimization problem with a genetic algorithm. Predictive control is developed for local optimization to refine the global optimization in real time, more particularly, the train operation modes including traction, cruise, coast, and braking are switched on the base of the line slope information, from which a more energy-efficient speed trajectory is generated under the constraints of fixed time and distance. To verify the effectiveness of the proposed strategy, operation of CHR-3 on high-speed railway is tested. Through the comparison of energy consumption in two typical cases, it verifies that the proposed energy-saving strategy works better than that of single optimization strategy. At the same time, the actual speed deviation can be corrected in a timely manner with the proposed method.

This paper considers a downlink massive MIMO frequency division duplexing (FDD) system. Due to the large number of antennas, the required length of training sequence for downlink training significantly increases in FDD mode, which leads to prohibitive overhead in real system. Thus, in this work we investigate how the length of training sequence affects the system performance. For this purpose, we derive an analytical expression of the ergodic achievable rate from a worst case viewpoint with the the training sequence length as a parameter in it. It is revealed from the analytical results that i.) the length of training sequence divided by the number of base station antennas approaches to zero yet the achievable rate can increase to infinity as long as the antenna number is sufficient large; ii.) there is a ceiling effect on the achievable rate if the antenna number grows large with any fixed training length. Furthermore, we propose a guideline for the selection of the training length. Numerical results validate the derivations and analysis.

In this paper, a massive MIMO downlink frequency division duplexing (FDD) system over correlated Rayleigh fading channel is considered. It is well known that the length of training sequence not only affects the accuracy of channel estimation but also accounts for the rate loss resulting from training overhead. However, as the number of the base station antennas becomes large, the required length of training sequence cannot increase unlimitedly. Thus, in this work we derive the analytical expression of achievable rate and investigate the impacts of the training sequence length on system asymptotic performance. It is discovered from the analytical results in two-fold that (1) the length of training sequence normalized by the antenna number approaches to zero yet the system capacity is guaranteed to positive infinity as long as the antenna number is large enough; (2) the transmission capability saturates to a certain level if the antenna number grows to very large with any given training length. Simulation results verify the theoretical derivations and demonstrate the performance limit.

This paper investigates the resource allocation issue between downlink training stage and data transmission stage for the frequency division duplexing (FDD) massive multiple-input multiple-output system from the viewpoint of energy efficiency (EE). For a given total energy budget during a coherence period, how to jointly select the training duration, training power and data power is of great significance for the system EE. Thus, an optimization problem of energy-efficient resource allocation is put forward. Since the analytical expression of the involved average spectral efficiency (SE) is intractable, a closed-form approximation of the SE is deduced using deterministic equivalent. Based on the simplified expression, the original non-convex fractional optimization problem is transformed into an equivalent problem in subtractive form by the means of fraction programming, which includes an achievable solution. Then, an iterative algorithm is proposed. Numerical results validates the benefits of the proposed resource allocation scheme.