Yanlong Zhao

Self-sustainable RIS can harvest energy to maintain its operation by adaptively adjusting its reflection coefficient matrix, which brings convenience in deployment and maintenance. In this letter, we consider a self-sustainable RIS-assisted multi-input multi-output simultaneous wireless information and power transfer (SWIPT) system. Our aim is to maximize the weighted sum power received by energy users while considering the minimum data rate requirements of information users and the minimum energy demand of RIS. To tackle the non-convex and highly coupled sum power maximization problem, we first employ minimum-mean-square error (MMSE) method to reformulate the original problem. Subsequently, we propose a block coordinate descent-based framework that decouples the reformulated problem into two subproblems which can be solved iteratively. In addition, successive convex approximation (SCA) method is adopted to approximate the non-convex objective function and constraint. Finally, simulation results confirm that the self-sustainable RIS can significantly improve the energy transmission performance.

Artificial intelligence technology, as a highly efficient auxiliary tool, is driving innovation in higher education teaching models. This paper focuses on the deep integration and application of AI technology in the experimental course of electronic circuit design. Addressing core issues such as the highly theoretical nature of the course, the abstract and difficult-to-understand components, and the insufficient development of students’ practical abilities, this study constructs and implements a new AI-integrated teaching model for the experimental course. By leveraging AI-assisted component recognition, knowledge graph construction, and learning analytics, this model comprehensively restructures the entire teaching process, including pre-class preparation, in-class instruction, experiment design, and teaching reflection. Empirical evidence shows that the introduction of AI tools significantly reduces the learning difficulty for students, provides a personalized and exploratory learning experience, enhances their understanding of circuit principles, and strengthens their capabilities in circuit design and fault diagnosis. This approach offers a new pathway and paradigm for the reform of basic electronic circuits laboratory courses.

The external power supply requirement of intelligent reflecting surfaces (IRS) restricts its deployment and application. To address this issue, self-sustainable IRS that can harvest energy from wireless signals significantly enhances the convenience and feasibility of IRS deployment. This paper investigates a novel IRS-aided simultaneous wireless information and power transfer (SWIPT) system, where IRS adopts power splitting (PS) protocol, time switching (TS) protocol or element splitting (ES) protocol to extract energy from the incident signals for sustaining its operation. Our aim is to maximize the weighted sum rate (WSR) while accounting for the constraints in terms of maximum transmit power, IRS reflection coefficients, the energy harvesting requirements of both users and the IRS. The WSR maximization problem is non-convex due to the coupling among variables. For the formulated problem, we propose an alternating optimization (AO) framework to transform it into several subproblems and solves them iteratively. Specifically, we utilize successive convex approximation (SCA) technique to tackle the non-convex optimization challenges in both transmit and reflective beamforming subproblems. Finally, simulation results demonstrate that employing a self-sustainable IRS together with the proposed algorithm yields a WSR improvement of 32%–60%.

In the domain of deep learning for visual tasks, ensemble models combine several less accurate models to form a more precise composite model, improving overall performance. Traditionally, majority voting and average probabilities have been the main decision-making techniques in ensemble learning, focusing only on the TOP1 Class of base models, hence overlooking other significant information. This article introduces a new algorithm, TOP2 hybrid decision (TOP2 HD), which enhances the TOP1 accuracy of the ensemble model. TOP2 HD categorizes base models into hierarchies based on their TOP1 Class and uses the TOP2 Class for ranking, leading to better performance. Extensive experiments across various models and datasets demonstrate that TOP2 HD not only surpasses traditional ensemble methods, such as majority voting, average probabilities, and stacking, but also exceeds many of the latest ensemble strategies in the image domain. In addition, our experiments revealed a functional relationship between the test accuracy of the ensemble model and the number of base models. This enables us to predict the upper limit of the ensemble model’s performance using only a fraction of the models, providing a crucial reference for the performance after the deployment of the ensemble model.

Pest detection in agriculture faces the challenge of adapting to new pest species while preserving the ability to recognize previously learned ones. Traditional model fine-tuning approaches often result in catastrophic forgetting, where the acquisition of new classes significantly impairs the recognition performance of existing ones. Although knowledge distillation has been shown to effectively mitigate catastrophic forgetting, current research predominantly focuses on feature imitation, neglecting the extraction of potentially valuable information from responses. To address this issue, we introduce a response-based distillation method, called adaptive response distillation (ARD). ARD incorporates an adaptive response filtering strategy that dynamically adjusts the weights of classification and regression responses based on the significance of the information. This approach selectively filters and transfers valuable response data, ensuring efficient propagation of category and localization information. Our method effectively reduces catastrophic forgetting during incremental learning, enabling the student detector to maintain memory of old classes while assimilating new pest categories. Experimental evaluations on the large-scale IP102 pest dataset demonstrate that the proposed ARD method consistently outperforms existing state-of-the-art algorithms across various class-incremental learning scenarios, significantly narrowing the performance gap compared to fully trained models.

Ultra-wideband (UWB) is increasingly popular in location-based services. Waveform division multiple access (WDMA) UWB utilizes orthogonal waveforms to achieve multiple access, enhancing system capacity and simplifying implementation. To address non-line-of-sight (NLOS) ranging errors, we propose a channel classification-based NLOS mitigation approach. Channels are categorized into line-of-sight, soft NLOS, and hard NLOS based on direct path attenuation levels. Our mitigation approach includes channel classification, ranging estimation, and positioning. Experiments were conducted under the IEEE 802.15.4a standard. We employ a squeeze and extraction convolutional neural network for dynamic environment channel classification. We propose an adaptive particle filter adjusting noise standard deviation based on channel classification outcomes to mitigate NLOS effects in ranging data. LSE-guided Quasi-Newton positioning method is combined with an anchor selection algorithm to further improve accuracy. Comparison experiments demonstrate the efficacy of our mitigation approach across diverse environments. In scenarios with a substantial amount of range measurements affected by NLOS, the positioning error has decreased 75%.

Unmanned aerial vehicles (UAVs) have played an integral part in emergency communications due to their excellent maneuverability. In this paper, we propose an intelligent reflecting surface (IRS)-assisted UAV multipair communication system, in which the IRS establishes a communication link between each pair of base stations (BSs) and users. Our goal is to maximize the minimum transmission rate by using relaxation and successive convex approximation (SCA) techniques. The optimization problem is nonconvex and difficult to address due to the presence of binary and coupled variables. Therefore, we relax the binary variables into continuous variables and decompose the problem into three subproblems for iterative optimization, leading to the derivation of the optimal solution. Initially, we address the optimization of resource allocation and IRS phase shift variables. Subsequently, we apply the SCA technique to resolve the UAV trajectory optimization problem. Finally, the above three subproblems are iterated until the algorithm converges. Numerical results indicate that the minimum transmission rate is effectively enhanced by employing the proposed algorithm.

Cognitive radio network (CRN) is highly vulnerable to malicious attacks, which can significantly impact the effectiveness and reliability of spectrum sensing. Existing deep learning-based works have proven ineffective in accurately recognizing an unknown attacker due to the prior information about the attacker for training purposes is not easily available in advance. To address the challenge of reliably identifying the known primary user (PU) and the unknown attacker, we combine deep learning with extreme value theory (EVT) to propose a reliable spectrum sensing method for primary user emulation attack (PUEA) detection. Importantly, our method does not rely on acquiring the prior information about the attacker during the training phase. The proposed deep learning-based method not only enables the secondary user (SU) to accurately confirm the presence of PU but also effectively identifies unknown attackers. In addition, the network structure of proposed method is redesigned to extract diverse multi-domain features. The simulation results indicate that the proposed method exhibits remarkable performance under the presence of unknown attacker.

The existing Automatic Modulation Classification (AMC) methods require the training and testing datasets share a common set of modulation categories. However, the AMC model may encounter the need to discriminate novel classes in non-cooperative environments. To the best of our knowledge, the research reporting AMC in the class-disjoint environment has not been addressed yet. In this letter, a novel class discovery method is proposed for AMC leveraging the information contained in the labeled training dataset. Specifically, a 3-stage deep learning method is introduced to recognize samples of the known classes and cluster samples of novel classes. The extracted features and the pairwise similarity relationship are considered as the common knowledge between the two class-disjoint datasets and are utilized to help the construction and training of the classifier for novel classes. The simulation results validate the effectiveness and performance of the proposed method.