The rolling mill guiding system is a key component that affects the quality of steel products. However, due to the harsh on-site environment, there is usually a lack of effective online monitoring and early warning mechanisms. Moreover, in industrial environments, fault samples are very scarce, making supervised artificial intelligence methods difficult to apply. This paper proposes a “physics-enhanced” orthogonal-sensing cyber-physical architecture that integrates hardware and software design. At the hardware level, an embedded orthogonal sensing layout () is designed to decouple drive-chain vibration from rolling-force fluctuations at the transducer level. At the algorithm level, the state monitoring of the guiding system is formulated as a self-supervised anomaly detection problem, and a two-branch network architecture is designed: one branch uses the CSD transformer to capture physical coupling characteristics, while the other branch uses VQ-VAE to extract operating-condition context. Experimental results on a dataset comprising real operational data and expert-validated synthetic fault scenarios show that the system achieves an AUC-ROC of 0.952 and a false alarm rate of 0.048 under a 95% TPR, with an end-to-end processing latency of approximately 8 ms per window and a system-level fault response time of approximately 108 ms, and thus meets the requirements of real-time industrial monitoring.

Today, access control systems are used in almost every institution and building. This is because they are an effective solution that provides a high level of security. There are many commercially available systems that provide security-related access features for buildings, including biometric options. Most use a centralized architecture, where each building can be remotely controlled via an Internet connection. This paper presents a completely different system from those on the market, a decentralized system with clone-detection and data-integrity verification mechanisms that allows access to buildings. The overall architecture includes hardware encoding of the access system’s location, and access is granted based on information written to the RFID card by the card-issuing center. This allows the system to be easily reconfigured at the hardware level prior to installation in the access area. The proposed system uses a confidential RFID card data integrity algorithm that uses the card data and immutable UID to determine a checksum in order to validate the RFID card data. As a result, any unwanted modification of at least one bit invalidates the card and blocks access to the building. The system was implemented, validated, and extensively tested over a one-year period with no reported operational issues.

A compact millimeter-wave radar system operating at 33 GHz is presented for integration on small unmanned aerial systems (UAS) and for ground-based counter-UAS reconnaissance. The design is specifically motivated by civil-sector agricultural applications, where large-payload crop-dusting and precision-spraying drones operating under FAA 14 CFR Part 137 require lightweight sense-and-avoid radar that conforms aerodynamically to existing aircraft or ground vehicles. The system is based on a 36-element hemispherical conformal phased array of crossed half-wave dipole radiators that generate right-hand circular polarization (RHCP) on transmit and selectively receives left-hand circular polarization (LHCP) echoes from targets, providing passive first-stage suppression of co-polarized rain and ground clutter. A Linearly Constrained Minimum Variance (LCMV) digital beamformer, applied to per-element analog-to-digital converter (ADC) outputs, delivers closed-form beam weights that enforce a distortionless response at each scan direction while globally minimizing sidelobe power. The formulation resolves the main-beam drift caused by the ill-conditioned re-scaling step in iterative Chebyshev tapering, achieving sidelobe levels below dB with main-beam peaks within ° of their commanded angles across all evaluated positions. Mutual coupling between array elements is modeled analytically using the induced-EMF method, yielding a impedance matrix whose off-diagonal entries are at most 8.2% of the element self-impedance at the minimum inter-element separation of 2.70 . A closed-form decoupling matrix is applied to the receive manifold prior to LCMV weight computation. Seven simultaneous independent receive beams covering °–° elevation are formed from a single data snapshot. A Scaled Conjugate Gradient neural network classifier, trained on radar-equation-scaled micro-Doppler features following Swerling I–IV radar cross-section (RCS) fluctuation statistics, achieves overall classification accuracy above 85% across five target classes. The five classes comprise two bird-signature classes (SW-I and SW-II), two UAV-signature classes (SW-III and SW-IV), and a clutter class. The design is entirely simulation-based; experimental validation using a sub-array prototype is identified as the primary direction for future work.

Accurate estimation of dynamic environmental phenomena through intelligent sensing systems plays a critical role in enabling reliable monitoring and decision-making in complex real-world scenarios. With the rapid development of artificial intelligence-driven sensing technologies and Internet of Things systems, modern agricultural monitoring is evolving from isolated data acquisition toward intelligent, multimodal perception and decision-making. However, traditional approaches predominantly rely on single data sources, making it difficult to simultaneously capture plant phenotypic variations and environment-driven mechanisms, thereby limiting model applicability in complex field scenarios. To address this issue, a multimodal pest density estimation framework, namely the Pest Density Estimation Framework (PDEF), is proposed, which integrates UAV-based imagery, trap monitoring data, and environmental sensor measurements. In this framework, crop canopy damage features are extracted using convolutional neural networks, while temporal encoding is employed to model dynamic environmental variations. Cross-modal feature alignment and environment-aware enhancement mechanisms are further introduced to achieve deep integration of multi-source information, enabling the construction of a unified feature representation space and improving estimation accuracy. Extensive experiments conducted on a constructed multimodal agricultural dataset demonstrate that the proposed method achieves MAE, RMSE, and MAPE values of , , and , respectively, significantly outperforming the Transformer-based fusion model (MAE , RMSE ). Meanwhile, the coefficient of determination reaches , indicating superior fitting capability and stability. In multimodal combination experiments, the three-modality fusion reduces error metrics by more than on average compared with single-modality models, validating the effectiveness of multi-source collaborative modeling. From the perspective of integrating plant phenotypic analysis and environmental perception, this study provides a novel AI-driven intelligent sensing framework for pest monitoring and crop management, contributing to improved pest prediction capability and enhanced intelligence in agricultural production systems. This study further provides practical implications for agricultural economics and supply chain optimization by enabling data-driven decision-making through intelligent sensing systems.

HighlightsWhat is the main finding?Proposed a spatiotemporal prediction and multi-modal fusion framework that actively freezes template updates (ηk ≈ 0) to reject specular reflections.What is the implications of the main finding?Achieved a 35% tracking error reduction and a 72.83 ms rapid re-identification latency, offering a highly robust solution for robotic visual sensing in complex environments.Rear-view human tracking and re-identification remain critical challenges for robotic visual sensing in unmanned vehicles, particularly under adverse weather conditions and severe occlusion. Conventional deep learning models often suffer from feature contamination and trajectory drift under dynamic illumination. To overcome these bottlenecks, we propose a lightweight tracking framework driven by spatiotemporal prediction and multimodal feature fusion. Specifically, an ego-motion-aware Kalman prediction mechanism maintains temporal continuity during complete occlusions. Upon target reappearance, a multi-factor descriptor—fusing color histograms with geometric constraints—is employed within a dynamic Mahalanobis search region. This is coupled with a specular-reflection-penalized adaptive learning rate (ηk) that actively freezes template updates during severe environmental degradation conditions. Evaluated on a custom Mecanum-wheeled robot, the proposed method achieves a peak precision of 94.2% and a tracking success rate of 93.4%. Extensive experiments in extreme rainy night scenarios demonstrate a 35% reduction in average tracking error, maintaining a Center Location Error (CLE) below 11 pixels. Furthermore, the system achieves a rapid target re-identification response of 72.83 ms during occlusion phases. Ultimately, this framework delivers a highly robust and real-time solution for autonomous navigation in complex dynamic environments.

In roll-to-roll (R2R) web processing systems, traction rollers impose precise velocity profiles on the moving web. Ideally, the web follows this trajectory without deviation, but slip can occur during rapid acceleration or deceleration, leading to tension loss and degraded product quality. Although slip can be detected directly using high-resolution encoders that track the actual web speed, such sensors are expensive and require machine downtime for installation, making them impractical for large-scale industrial deployment. To overcome this limitation, we developed a virtual slip sensor that estimates slip using existing machine signals only. A temporary encoder was used to collect ground-truth data, enabling the training of predictive models that eliminate the need for a permanent physical sensor. The proposed system employs an ensemble modeling approach: a CatBoost model captures low-slip behavior where data is abundant, while a linear model extrapolates to high-slip, out-of-distribution conditions. Targeted feature engineering ensures generalization across varying ramp times and web speeds. Despite being trained primarily on data containing limited slip, the models successfully generalized to scenarios with severe slip, demonstrating robust predictive performance. The ensemble reduces the regular CatBoost model’s MSE at 60 m/min by approximately 54% in the speed-based evaluation and by approximately 68% in the quantile-based evaluation while maintaining comparable performance in the low-speed regimes. The resulting virtual sensor enables continuous real-time slip monitoring, providing operators with timely insights to prevent quality degradation and operate at higher acceleration profiles to increase throughput, even on machines that have not previously experienced extreme slip.

Reconfigurable intelligent surface sensors (RISs) have emerged as a promising technology for enhancing wireless indoor localization by intelligently controlling signal propagation; however, extracting reliable localization fingerprints from RIS-assisted signals remains challenging due to multipath fading, environmental noise, and nonlinear spatial–temporal channel dynamics. To address this, we propose an Adaptive Dual-Reinforcement Learning-Hybrid Spatial–Temporal Network (ADRL-HSTNet) for RIS-assisted indoor localization. The framework utilizes dual-channel RSSI and phase measurements, followed by noise filtering, normalization, and sliding-window segmentation prior to feature extraction. It then constructs enhanced representations through handcrafted feature extraction and multi-branch processing, including patch-based features, wavelet-domain representations, statistical descriptors, and multi-level segmentation masks. These heterogeneous inputs are encoded using lightweight transformer-based encoders to capture multiscale dependencies. A first reinforcement learning selector adaptively weights the most informative feature branches to produce a fused representation, which is further processed by spatial and temporal transformer modules. Their outputs are adaptively combined via a second reinforcement learning selector to obtain robust localization embedding. The model jointly performs classification, coordinate regression, and uncertainty estimation end-to-end. Experimental results across multiple RIS configurations outperformed the KAN, LSTM-KAN, and RHL-Net (compared against the proposed ADRL-HSTNet) baselines, achieving accuracies of 83.33%, 75.22%, 93.33%, and 88.89%, confirming the effectiveness of the proposed approach.

Image demosaicking aims to reconstruct a full-resolution color image from spatially sparse and interleaved color filter array observations. Despite the significant progress achieved by deep learning-based methods, existing approaches have not fully addressed the sampling-structure-constrained nature of demosaicking. In particular, four-channel half-resolution packing may disrupt the CFA spatial phase relationships, while local convolutions and global non-local matching struggle to model reconstruction-relevant cross-position dependencies. To address these issues, this paper proposes an end-to-end image demosaicking network with region-level non-local modeling and residual aggregation (RNRA-Net). Instead of packing Bayer RAW data into a four-channel half-resolution representation, RNRA-Net decomposes the original mosaic image into a three-channel representation at the original resolution, thereby preserving the spatial arrangement of CFA sampling. To capture structurally related information, a region-level non-local module is introduced to compute feature correlations within spatially bounded regions, enabling aggregation of reconstruction-relevant contextual information. In addition, a residual aggregation module is developed to explicitly collect and refine early residual compensation features, facilitating the recovery of edges, textures, and high-frequency details. Extensive experiments on benchmark and high-resolution datasets demonstrate the effectiveness of RNRA-Net.

This paper proposes a digital potentiometer-based adaptive gas sensor interface for stable detection without signal saturation under extreme environmental fluctuations. Conventional fixed-gain circuits often suffer from limited dynamic range, leading to data loss when severe baseline drifts exceed ADC input limits. To address this, we developed a real-time control algorithm that actively adjusts attenuator and amplifier gains, maintaining the ADC input voltage (VADC) near the common-mode voltage (VCM). Experimental results demonstrate that the interface remains stable even when the buffer voltage reaches 2.75 V, significantly surpassing the 1.2 V ADC limit. Sensor resistance data, reconstructed by inversely calculating updated circuit parameters, achieved high accuracy with a Mean Absolute Percentage Error (MAPE) of 1.628% and a maximum relative error under 4.8%. Consequently, this study proves that logically extending the physically limited ADC dynamic range enables high-precision gas sensing in diverse environments without requiring high-performance computing devices. This approach provides a cost-effective and robust solution for compact IoT-based gas monitoring systems.

Machine learning (ML) has been increasingly considered for various communication applications, demonstrating promising feasibility and effectiveness in enhancing system intelligence, adaptability, and operational efficiency [...].