Sensors

23 pages, 4371 KB

Open AccessArticle

A Deep Learning-Based Method for Stress Measurement Using Longitudinal Critically Refracted Waves

by Yong Gan, Jingkun Ma, Binpeng Zhang, Yang Zheng, Xuedong Wang, Yuhong Zhu, Yibo Wang and Dachun Ji

Sensors 2026, 26(7), 2283; https://doi.org/10.3390/s26072283 - 7 Apr 2026

Viewed by 669

Abstract

Accurate stress measurement is essential to evaluating structural integrity and plays a pivotal role in the health monitoring and predicting the service life of steel infrastructures. This study proposes a deep learning approach for stress prediction based on longitudinal critically refracted (LCR) ultrasonic [...] Read more.

Accurate stress measurement is essential to evaluating structural integrity and plays a pivotal role in the health monitoring and predicting the service life of steel infrastructures. This study proposes a deep learning approach for stress prediction based on longitudinal critically refracted (LCR) ultrasonic waves. The model integrates gated recurrent units (GRU), attention mechanisms, and one-dimensional convolutional neural networks (1D-CNN), enabling direct stress prediction from raw ultrasonic signals without the need for manual feature extraction or explicit physical modeling. To validate the approach, LCR signals were acquired using a custom-built piezoelectric ultrasonic system from 20# steel specimens subjected to uniaxial stresses ranging from 0 to 200 MPa. A dataset comprising 4200 samples was augmented to enhance training efficiency. The proposed model achieved a mean absolute error of 1.94 MPa. Generalization tests demonstrated high accuracy across diverse stress levels, with average errors below 3 MPa, highlighting the model’s robustness. This research presents an accurate, intelligent, and calibration-free ultrasonic method for stress evaluation, providing practical support for stress evaluation in steel structures under actual operating conditions. Full article

(This article belongs to the Section Intelligent Sensors)

► Show Figures

Figure 1

27 pages, 4289 KB

Open AccessArticle

Online Extrinsic Calibration of Camera and LiDAR Based on Cascade Optimization

by Chuanxun Hou, Zheng He, Tong Zhao, Zhenhang Guo and Xinchun Ji

Sensors 2026, 26(7), 2282; https://doi.org/10.3390/s26072282 - 7 Apr 2026

Cited by 1 | Viewed by 994

Abstract

Accurate and stable extrinsic calibration is the foundation of high-quality fusion sensing and positioning of camera and Light Detection and Ranging (LiDAR). However, traditional targetless calibration methods suffer from limitations such as poor scene adaptability and unstable convergence, which significantly restrict calibration accuracy [...] Read more.

Accurate and stable extrinsic calibration is the foundation of high-quality fusion sensing and positioning of camera and Light Detection and Ranging (LiDAR). However, traditional targetless calibration methods suffer from limitations such as poor scene adaptability and unstable convergence, which significantly restrict calibration accuracy and robustness in complex environments. Aiming at solving those problems, we propose an online cascade-optimization-based extrinsic calibration method of combining motion trajectory alignment and edge feature alignment. In the initial calibration stage, a hand–eye calibration algorithm is designed by minimizing the residual discrepancies between camera odometry and LiDAR odometry sequences. It establishes a robust initialization for subsequent optimization. Then, in order to extract robust edge line features from sparse point clouds, we employ depth difference and planar edges of point clouds in the optimization process. Subsequently, principal component analysis (PCA) is applied to compute the principal direction of the extracted line features, enabling a decoupled optimization scheme that accounts for directional observability. This approach effectively mitigates the adverse effects of uneven environmental feature distributions. Experimental validation on typical urban datasets demonstrates the method’s generalizability and competitive accuracy: rotational parameter errors are constrained within 0.25°, and translational errors are maintained below 0.05 m. This affirms the method’s suitability for high-accuracy engineering applications. Full article

(This article belongs to the Special Issue Intelligent Sensor Calibration: Techniques, Devices and Methodologies)

► Show Figures

Figure 1

16 pages, 1624 KB

Open AccessArticle

Surface EMG-Based Hand Gesture Recognition Using a Hybrid Multistream Deep Learning Architecture

by Yusuf Çelik and Umit Can

Sensors 2026, 26(7), 2281; https://doi.org/10.3390/s26072281 - 7 Apr 2026

Cited by 1 | Viewed by 1298

Abstract

Surface electromyography (sEMG) enables non-invasive measurement of muscle activity for applications such as human–machine interaction, rehabilitation, and prosthesis control. However, high noise levels, inter-subject variability, and the complex nature of muscle activation hinder robust gesture classification. This study proposes a multistream hybrid deep-learning [...] Read more.

Surface electromyography (sEMG) enables non-invasive measurement of muscle activity for applications such as human–machine interaction, rehabilitation, and prosthesis control. However, high noise levels, inter-subject variability, and the complex nature of muscle activation hinder robust gesture classification. This study proposes a multistream hybrid deep-learning architecture for the FORS-EMG dataset to address these challenges. The model integrates Temporal Convolutional Networks (TCN), depthwise separable convolutions, bidirectional Long Short-Term Memory (LSTM)–Gated Recurrent Unit (GRU) layers, and a Transformer encoder to capture complementary temporal and spectral patterns, and an ArcFace-based classifier to enhance class separability. We evaluate the approach under three protocols: subject-wise, random split without augmentation, and random split with augmentation. In the augmented random-split setting, the model attains 96.4% accuracy, surpassing previously reported values. In the subject-wise setting, accuracy is 74%, revealing limited cross-user generalization. The results demonstrate the method’s high performance and highlight the impact of data-partition strategies for real-world sEMG-based gesture recognition. Full article

(This article belongs to the Special Issue Machine Learning in Biomedical Signal Processing)

► Show Figures

Figure 1

19 pages, 2676 KB

Open AccessArticle

Periodically Pulsed Polarization Gas Sensors Based on Au|YSZ: Mechanism of NO_x Detection

by Nils Donker, Jens Zosel, Ralf Moos and Daniela Schönauer-Kamin

Sensors 2026, 26(7), 2280; https://doi.org/10.3390/s26072280 - 7 Apr 2026

Viewed by 643

Abstract

Pulsed polarization of Au|YSZ gas sensors is examined to clarify the mechanism of NO_x detection under dynamic operation and to disentangle catalytic surface effects from electrochemical relaxation. Using gold electrodes with substantially lower catalytic activity than platinum explicitly enables this mechanistic separation. [...] Read more.

Pulsed polarization of Au|YSZ gas sensors is examined to clarify the mechanism of NO_x detection under dynamic operation and to disentangle catalytic surface effects from electrochemical relaxation. Using gold electrodes with substantially lower catalytic activity than platinum explicitly enables this mechanistic separation. During pulsed polarization, periodic voltage pulses are followed by self-discharge under open-circuit conditions, and the response is measured based on the self-discharge rate. NO₂ consistently accelerates the self-discharge from the beginning, whereas NO slows the relaxation predominantly at later times. CO and H₂ produce similar delaying effects, and C₃H₆ shows no measurable influence under the tested conditions. Decreasing ambient O₂ slows the discharge and amplifies the NO₂ effect, which indicates that oxygen supply and surface exchange at the triple-phase boundary are rate determining. A Pt-containing catalytic overlayer drives local NO/NO₂ interconversion toward equilibrium so that both gases yield to an accelerated self-discharge. These findings support a mechanistic picture in which NO₂ provides effective oxygen equivalents that accelerate discharge, whereas NO, CO, and H₂ consume oxygen and slow down discharge. Overall, this establishes a materials-based approach for distinguishing between NO and NO₂ and evaluating the underlying mechanism during pulsed polarization. Full article

(This article belongs to the Special Issue Recent Advances in Sensors for Chemical Detection Applications (2nd Edition))

► Show Figures

Figure 1

20 pages, 2963 KB

Open AccessArticle

Research on a Miniaturized Digital Servo System for Passive Hydrogen Masers

by Siyuan Guo, Meng Cao, Pengfei Chen, Tao Shuai, Wangwang Hu and Yuxian Pei

Sensors 2026, 26(7), 2279; https://doi.org/10.3390/s26072279 - 7 Apr 2026

Viewed by 520

Abstract

High-precision time and frequency references are essential for satellite navigation, deep-space exploration, and space science missions. To address the large size, high power consumption, and limited integration of conventional Passive Hydrogen Maser (PHM) servo electronics based on discrete analog chains, this paper proposes [...] Read more.

High-precision time and frequency references are essential for satellite navigation, deep-space exploration, and space science missions. To address the large size, high power consumption, and limited integration of conventional Passive Hydrogen Maser (PHM) servo electronics based on discrete analog chains, this paper proposes a miniaturized digital servo architecture for PHMs based on software-defined radio (SDR) and a field-programmable gate array (FPGA). The AD9364 is used as an integrated RF front end for microwave interrogation signal generation, receiver down-conversion, and analog-to-digital conversion (ADC), while digital demodulation, discriminator construction, and closed-loop control are implemented in the FPGA. A dual-frequency interrogation and time-division multiplexing scheme is introduced to separate the atomic and cavity responses, and an oversampling-based processing method combining outlier rejection and averaging decimation is adopted to improve the observation accuracy and noise immunity of weak error signals. Experimental results demonstrate stable closed-loop locking of the atomic transition spectrum, achieving a frequency stability of 1.46 × 10⁻¹² at 1 s, while significantly improving the compactness and integration level of the servo electronics. Full article

(This article belongs to the Section Navigation and Positioning)

► Show Figures

Figure 1

36 pages, 6596 KB

Open AccessArticle

Co-Design of Smartphone- and Smartwatch-Based Occupational Health Visualisations in Office Environments

by Phillip Probst, Sara Santos, Gonçalo Barros, Mariana Morais, Sofia Garcia, Philipp Koch, Jorge Barroso Dias, Ana Leal, Rute Periquito, Sofia André, Tiago Matoso, Cristina Pinho, Ricardo Vigário and Hugo Gamboa

Sensors 2026, 26(7), 2278; https://doi.org/10.3390/s26072278 - 7 Apr 2026

Viewed by 848

Abstract

Office workers are exposed to a range of occupational health risks, including prolonged sedentary behaviour, postural load, elevated heart rate, and noise, yet objective and continuous monitoring of these risk factors in workplace settings remains uncommon. This study aimed to co-design occupational health [...] Read more.

Office workers are exposed to a range of occupational health risks, including prolonged sedentary behaviour, postural load, elevated heart rate, and noise, yet objective and continuous monitoring of these risk factors in workplace settings remains uncommon. This study aimed to co-design occupational health visualisations based on smartphone and smartwatch data, through a multi-stakeholder group of office workers and occupational health professionals. A generative co-design framework was applied, comprising a pre-design phase with a field study and questionnaire, a structured multi-stakeholder workshop, and a follow-up evaluation session. Thematic analysis of the workshop transcript yielded 17 occupational health themes, which were subsequently assessed for technical feasibility relative to the available sensing platform. Of the 27 discrete visualisation elements proposed across both groups, the majority were classified as directly addressable using smartphone and smartwatch sensor data. Visualisations covering physical activity, heart rate, environmental noise exposure, and postural load were implemented in Python using real-world data collected from office workers. The follow-up session provided qualitative confirmation that the developed visualisations were interpretable and aligned with the stakeholder expectations. The generative co-design framework proved well-suited to the occupational health visualisation context, enabling structured translation of stakeholder requirements into technically feasible and interpretable visualisation outputs. Full article

(This article belongs to the Special Issue Recent Trends and Advances in Wearable Exoskeletons, Sensing Technologies and Occupational Ergonomics)

► Show Figures

Figure 1

30 pages, 3241 KB

Open AccessArticle

A Joint Framework of IMM-LSTM-C Tracking and IBPDO-Based Node Selection for Energy-Efficient Cooperative Tracking in Underwater Acoustic Sensor Networks

by Wenbo Zhang, Yadi Hou and Hongbo Zhu

Sensors 2026, 26(7), 2277; https://doi.org/10.3390/s26072277 - 7 Apr 2026

Viewed by 483

Abstract

The increasing deployment of underwater vehicles demands accurate and energy-efficient target tracking in sensor networks. However, existing approaches have largely addressed tracking accuracy and energy efficiency in isolation, and a system-level framework that jointly optimizes both remains lacking. To address this gap, this [...] Read more.

The increasing deployment of underwater vehicles demands accurate and energy-efficient target tracking in sensor networks. However, existing approaches have largely addressed tracking accuracy and energy efficiency in isolation, and a system-level framework that jointly optimizes both remains lacking. To address this gap, this paper proposes a joint optimization framework with two main contributions. First, to improve tracking accuracy under complex maneuvering conditions, we develop an Interactive Multi-Model using Long Short-Term Memory Classification (IMM-LSTM-C) algorithm, which integrates multi-step model likelihoods into an LSTM network for precise motion classification, achieving a 7.1% accuracy improvement over IMM-BP. Second, to reduce network energy consumption while maintaining tracking performance, we introduce an Improved Binary Prairie Dog Optimization (IBPDO) algorithm for node selection, enhanced with Cauchy mutation and opposition-based learning. Simulation results show that IBPDO achieves 6.1–8.2% higher accuracy than BWOA and reduces energy consumption by 12% compared to LNS. Furthermore, the complete joint framework demonstrates synergistic effects, reducing tracking error by 19.3% and energy consumption by 15.4% over the IMM + LNS baseline. The proposed framework provides an effective balance between tracking accuracy and energy efficiency in underwater acoustic sensor networks. Full article

(This article belongs to the Special Issue Innovative Protocols, Algorithms, and Applications for Autonomous Sensing and Sensor Networks)

► Show Figures

Figure 1

16 pages, 1212 KB

Open AccessArticle

Quad-Element Implantable MIMO Antenna for Wireless Capsule Endoscopy

by Amor Smida, Jun Jiat Tiang, Mohamed I. Waly and Surajo Muhammad

Sensors 2026, 26(7), 2276; https://doi.org/10.3390/s26072276 - 7 Apr 2026

Cited by 1 | Viewed by 700

Abstract

Compared to antennas bearing a single port, MIMO antennas with several ports enable higher data throughput by exploiting spatial diversity. This capability is essential for next-generation implantable medical devices, where high channel capacity is a key requirement. A quad-element implantable MIMO antenna is [...] Read more.

Compared to antennas bearing a single port, MIMO antennas with several ports enable higher data throughput by exploiting spatial diversity. This capability is essential for next-generation implantable medical devices, where high channel capacity is a key requirement. A quad-element implantable MIMO antenna is designed and practically validated at 1420 MHz in this paper. It occupies a compact volume of

7 \times 8 \times 0.1 {mm}^{3}

(

5.6 {mm}^{3}

). The compactness is realized by combining high-permittivity substrate (Rogers 3010 with relative permittivity of 10.2) with meandered radiator paths, which increase the effective current length while maintaining a small physical size. All antennas have very small mutual coupling with isolation of more than

31.78

dB, which is mainly due to the spacing of 1 mm between the elements and the substrate, which is thin. The peak realized gain for each antenna element is

- 27.3

dBi. The simulation is performed within a capsule-like structure, which is embedded in the stomach tissue model. The experimental verification is carried out by embedding antenna within minced meat. The ECC, channel capacity, and link margin are also evaluated and found to be satisfactory. The proposed antenna ensures reliable communication performance, with the transmission range being as high as

2.5

m, link margin being 15 dB, and the data rate being 120 Mb/s. The proposed antenna ensures a good level of ECC, which is less than

0.1

. The SAR is

52.3

W/kg at 1420 MHz. This design is favorable for implants because of the small size, good impedance matching, high isolation, low correlation, good level of gain, and good link performance. Full article

(This article belongs to the Section Biomedical Sensors)

► Show Figures

Figure 1

9 pages, 640 KB

Open AccessCommunication

Noninvasive Measurement of Infant Respiration During Sleep: A Validation Study

by Melissa N. Horger, Maristella Lucchini, Shambhavi Thakur, Rebecca M. C. Spencer and Natalie Barnett

Sensors 2026, 26(7), 2275; https://doi.org/10.3390/s26072275 - 7 Apr 2026

Viewed by 858

Abstract

Infant respiration is a physiological marker of health and wellbeing that can provide insight into sleep and wake patterns. Technological innovation presents opportunities to enhance measurements of physiological signals, which improves ecological validity and participant experiences. This is particularly true in the context [...] Read more.

Infant respiration is a physiological marker of health and wellbeing that can provide insight into sleep and wake patterns. Technological innovation presents opportunities to enhance measurements of physiological signals, which improves ecological validity and participant experiences. This is particularly true in the context of studying infant sleep, as it can be disrupted by changes in the environment and the physical sensation of unfamiliar or uncomfortable sensors. The goal of this study was to examine if a commercially available video baby monitor (Nanit system) can accurately estimate respiration during a nap relative to a commonly used cardiorespiratory sensor (Isansys Lifetouch sensor). Thirty-three infants (M = 9.7 months; range = 1–22 months) took a nap while wearing the Lifetouch sensor and Nanit Breathing Band. Infants slept in view of the Nanit camera. A computer vision algorithm applied to the video detected movement of the patterns on the fabric band worn around the infant’s torso to determine respiratory rates. The results showed strong consistency between the devices. More than 95% of the minute-by-minute respiration data fell within the limits of agreement, with little bias. Agreement was not influenced by age or nap duration, suggesting the Nanit Breathing Band provides a valid measure of respiration across infancy. Full article

(This article belongs to the Collection Biomedical Imaging and Sensing)

► Show Figures

Figure 1

17 pages, 2678 KB

Open AccessArticle

A Novel Workflow to Estimate Limb Orientation from Wearable Sensors to Monitor Infant Motor Development

by David Song, William J. Kaiser, Sitaram Vangala and Rujuta B. Wilson

Sensors 2026, 26(7), 2274; https://doi.org/10.3390/s26072274 - 7 Apr 2026

Viewed by 980

Abstract

Background: Wearable sensors have gained increasing popularity as an objective method for remotely monitoring infant movement in naturalistic settings. Over the first year of life, infants generate a wide range of motions, from goal-directed to spontaneous movement. These include linear movements, such as [...] Read more.

Background: Wearable sensors have gained increasing popularity as an objective method for remotely monitoring infant movement in naturalistic settings. Over the first year of life, infants generate a wide range of motions, from goal-directed to spontaneous movement. These include linear movements, such as kicks, and orientation changes, such as postural transitions. Many sensor processing pipelines emphasize capturing linear movements through movement-generated acceleration while focusing less on information about orientation embedded in the gravitational part of the data. Here, we introduce a complementary gravity-referenced approach that extracts the gravitational component of accelerometer signals to estimate limb orientation, extending the reliable quantification of rich and detailed aspects of infant movement. Infant orientation has demonstrated clinical relevance, including associations with later neuromotor outcomes, and it can be used to chart infant motor development, motivating the development of objective methods to quantify orientation from sensor data. Methods: Wearable sensors (Opal APDM) were used to longitudinally evaluate infant motor activity recorded in sessions conducted at 3, 6, 9, and 12 months of age. We extracted data from a 5 min segment that has simultaneous video recordings. From these datasets, applying the gravity-referenced method, we computed pitch, roll, and yaw, angles that collectively describe limb orientation. We then quantified orientation variability using axis-specific circular standard deviations (SDs) for pitch, roll, and yaw and a multi-axis composite measure based on generalized variance. Results: Axis-specific circular SDs for pitch, roll, and yaw, as well as the composite generalized variance, increased significantly from 3 to 12 months (p ≤ 0.01 for each metric). Composite variability was strongly associated with Mullen gross motor outcomes at 9 and 12 months of age (r = 0.55, p < 0.001). Conclusions: Overall, gravity-referenced pitch, roll, and yaw provide rich orientation features that increased as infants develop more postural transitions. Furthermore, the orientation features correlated with standardized measures of infant motor function. These orientation metrics can complement traditional linear kinematic measures and improve our ability to granularly track infant motor development in the first year of life. Full article

(This article belongs to the Section Wearables)

► Show Figures

Figure 1

25 pages, 2327 KB

Open AccessArticle

Joint Beamforming for Integrated Satellite–Terrestrial ISAC Systems

by Tengyu Wang and Qian Wang

Sensors 2026, 26(7), 2273; https://doi.org/10.3390/s26072273 - 7 Apr 2026

Viewed by 826

Abstract

Satellite–terrestrial integrated networks provide seamless global coverage, especially in remote areas where terrestrial deployment is costly. Integrated sensing and communications (ISAC) enhances spectral efficiency by merging both functions on a single platform. This paper proposes a novel integrated satellite–terrestrial ISAC architecture, where a [...] Read more.

Satellite–terrestrial integrated networks provide seamless global coverage, especially in remote areas where terrestrial deployment is costly. Integrated sensing and communications (ISAC) enhances spectral efficiency by merging both functions on a single platform. This paper proposes a novel integrated satellite–terrestrial ISAC architecture, where a satellite performs simultaneous communication and sensing. The satellite transmits communication signals and sensing waveforms to an Earth Station, which then relays them to a terrestrial base station to serve multiple users. We formulate a joint beamforming design problem to maximize the sum rate of users under quality-of-service constraints, backhaul capacity limits, beampattern requirements for sensing, and power budgets. With perfect channel state information, the non-convex problem is transformed into a difference-of-convex form and solved via the convex–concave procedure. For imperfect channel state information, a robust method combining successive convex approximation and the S-procedure is developed. Simulations show the proposed design outperforms benchmarks and is suitable for low-Earth orbit satellite systems. Full article

(This article belongs to the Special Issue New Technologies in Wireless Communication System)

► Show Figures

Figure 1

32 pages, 1006 KB

Open AccessSystematic Review

LEACH Protocol Evolution in WSN: A Review of Energy Consumption Optimization and Security Reinforcement

by Aijia Chu, Tianning Zhang and Chengyi Wang

Sensors 2026, 26(7), 2272; https://doi.org/10.3390/s26072272 - 7 Apr 2026

Cited by 3 | Viewed by 1368

Abstract

As a foundational protocol in wireless sensor networks (WSNs), LEACH has long contended with the dual challenges of energy load balancing and security defense. To clarify the protocol’s evolutionary trajectory within the modern IoT context, this paper presents a systematic review and restructuring [...] Read more.

As a foundational protocol in wireless sensor networks (WSNs), LEACH has long contended with the dual challenges of energy load balancing and security defense. To clarify the protocol’s evolutionary trajectory within the modern IoT context, this paper presents a systematic review and restructuring of LEACH’s optimization mechanisms. The core contributions of this study are threefold: First, it establishes a taxonomy for energy optimization in LEACH. It provides an in-depth analysis of how intelligent algorithms—such as fuzzy logic and meta-heuristics—reshape cluster head election and data transmission paths in heterogeneous network environments, thereby resolving the inherent blindness of traditional mechanisms. Second, it elucidates the evolutionary patterns of LEACH security mechanisms. The paper details the transition of defense strategies from early static encryption and authentication to dynamic countermeasure mechanisms, offering a clear framework for understanding the protocol’s defensive boundaries. Finally, addressing the bottleneck where high security levels often incur high energy costs, the paper explores the feasibility of incorporating zero-trust architecture (ZTA) into WSNs within the future outlook section. This discussion aims to provide a new theoretical perspective for future research on balancing enhanced defense capabilities with energy efficiency. Full article

(This article belongs to the Section Internet of Things)

► Show Figures

Figure 1

20 pages, 1234 KB

Open AccessArticle

Lightweight Real-Time Navigation for Autonomous Driving Using TinyML and Few-Shot Learning

by Wajahat Ali, Arshad Iqbal, Abdul Wadood, Herie Park and Byung O Kang

Sensors 2026, 26(7), 2271; https://doi.org/10.3390/s26072271 - 7 Apr 2026

Viewed by 871

Abstract

Autonomous vehicle navigation requires low-latency and energy-efficient machine learning models capable of operating in dynamic and resource-constrained environments. Conventional deep learning approaches are often unsuitable for real-time deployment on embedded edge devices due to their high computational and memory demands. In this work, [...] Read more.

Autonomous vehicle navigation requires low-latency and energy-efficient machine learning models capable of operating in dynamic and resource-constrained environments. Conventional deep learning approaches are often unsuitable for real-time deployment on embedded edge devices due to their high computational and memory demands. In this work, we propose a unified TinyML-optimized navigation framework that integrates a lightweight convolutional feature extractor (MobileNetV2) with a metric-based few-shot learning classifier to enable rapid adaptation to unseen driving scenarios with minimal data. The proposed framework jointly combines feature extraction, few-shot generalization, and edge-aware optimization into a single end-to-end pipeline designed specifically for real-time autonomous decision-making. Furthermore, post-training quantization and structured pruning are employed to significantly reduce the memory footprint and inference latency while preserving the classification performance. Experimental results demonstrate that the proposed model achieved a 93.4% accuracy on previously unseen road conditions, with an average inference latency of 68 ms and a memory usage of 18 MB, outperforming traditional CNN and LSTM models in efficiency while maintaining a competitive predictive performance. These results highlight the effectiveness of the proposed approach in enabling scalable, real-time navigation on low-power edge devices. Full article

(This article belongs to the Special Issue Scalable Blockchain and AI-Based Embedded IoT Systems for Smart Spaces (3rd Edition))

► Show Figures

Figure 1

19 pages, 12031 KB

Open AccessTechnical Note

Efficient Mesh Reconstruction and Texturing of Oracle Bones

by Shiming De

Sensors 2026, 26(7), 2270; https://doi.org/10.3390/s26072270 - 7 Apr 2026

Cited by 1 | Viewed by 578

Abstract

The high-fidelity 3D digitization of small, detailed cultural heritage objects, such as Oracle Bones, presents significant challenges for which existing reconstruction workflows are often inadequate. Methods based on Structure-from-Motion (SfM) often lack the geometric density required to capture fine inscription details, while Light [...] Read more.

The high-fidelity 3D digitization of small, detailed cultural heritage objects, such as Oracle Bones, presents significant challenges for which existing reconstruction workflows are often inadequate. Methods based on Structure-from-Motion (SfM) often lack the geometric density required to capture fine inscription details, while Light Detection and Ranging and RGB-Depth approaches may introduce high data overhead and unstable color mapping. Recent specialized studies have utilized multi-shading-based techniques to extract such hidden surface textures, yet integrating these results into a cohesive mesh remains difficult. To address these limitations, we propose a digitization framework specifically designed for object-level archaeological artifacts. Our method combines semi-automatic alignment with ICP-based refinement for robust camera pose estimation, reducing misalignment issues associated with feature-only registration. Furthermore, we employ an efficient mesh-based representation with vertex-level coloring, enabling detailed geometry and consistent texturing while maintaining compact storage requirements. Our contributions include: (1) a high-quality mesh reconstruction framework that preserves fine inscription geometry; (2) a hybrid camera pose estimation strategy that improves alignment robustness; and (3) an integrated hardware-assisted workflow tailored for digitizing small archaeological artifacts under controlled acquisition conditions. Experimental results on physical Oracle Bone artifacts demonstrate that the proposed method achieves a mean geometric reconstruction error of approximately 0.075 mm with a Hausdorff distance of 1 mm. These results demonstrate the effectiveness of the proposed workflow for digitization of oracle bone artifacts. Full article

(This article belongs to the Section Sensor Networks)

► Show Figures

Figure 1

21 pages, 2519 KB

Open AccessArticle

PyAO: PyTorch-Based Memory-Efficient LLM Training on Ethernet-Interconnected Clusters

by Daemin Kim, Hyorim Kim, Juncheol Ahn and Sejin Park

Sensors 2026, 26(7), 2269; https://doi.org/10.3390/s26072269 - 7 Apr 2026

Viewed by 953

Abstract

As large language models (LLMs) pursue higher accuracy, their model sizes have surged, substantially increasing GPU memory consumption. Prior work mitigates this issue by distributing the memory burden across multiple GPUs. However, on clusters interconnected via Ethernet, the resulting computational intensity is insufficient [...] Read more.

As large language models (LLMs) pursue higher accuracy, their model sizes have surged, substantially increasing GPU memory consumption. Prior work mitigates this issue by distributing the memory burden across multiple GPUs. However, on clusters interconnected via Ethernet, the resulting computational intensity is insufficient to hide the significant network latency. Achieving a favorable compute-to-communication ratio is further constrained by the memory required to cache the massive activations generated during the forward pass. PyAO, proposed in this paper, effectively offloads activations, selects offloading strategies based on their offloading efficiency, and minimizes data-movement bottlenecks, thereby enabling larger micro-batch sizes. In Ethernet-interconnected cluster environments, experiments on popular models—including OPT-1.3B, GPT-0.8B, and Llama-1.2B—demonstrate that PyAO reduces peak GPU memory by up to 1.94× at the same micro-batch size, enables up to 2.5× larger batch sizes, and accelerates training by up to 3.63× relative to the baseline. Full article

(This article belongs to the Special Issue Intelligent Sensors and Advanced Computing: Developments in the Era of Industry 4.0: 2nd Edition)

► Show Figures

Figure 1

16 pages, 2525 KB

Open AccessArticle

Novel Technology for Unbalance Diagnosis for Dual-Speed Wind Turbines

by Amir R. Askari, Len Gelman, Russell King, Daryl Hickey and Mehdi Behzad

Sensors 2026, 26(7), 2268; https://doi.org/10.3390/s26072268 - 7 Apr 2026

Viewed by 1476

Abstract

Unbalance diagnosis for non-constant speed systems is challenging because the 1X fundamental rotational harmonic magnitude, commonly used as an unbalance indicator, depends on shaft rotational speed. This dependency makes it difficult to separate speed effects from unbalance effects. It has been shown that [...] Read more.

Unbalance diagnosis for non-constant speed systems is challenging because the 1X fundamental rotational harmonic magnitude, commonly used as an unbalance indicator, depends on shaft rotational speed. This dependency makes it difficult to separate speed effects from unbalance effects. It has been shown that 1X magnitudes become speed-invariant if they are normalized with respect to the rotational speed in power four for variable-speed wind turbines. However, the applicability of this diagnostic technology to dual-speed machines remains unclear. This study experimentally investigates unbalance diagnosis technologies for dual-speed wind turbines, for which speed-dependent interference is present. Vibration data are collected from the main bearings of two dual-speed wind turbines. Novel residual-based, speed-invariant unbalance diagnostic technology is proposed. The experimental results show consistent statistical distributions of the new diagnosis indicator across low and high-speed operating regimes. These findings confirm the suitability of the proposed technology for unbalance diagnosis for dual-speed rotating machinery. Full article

(This article belongs to the Special Issue Intelligent Sensors for Structural Health Monitoring and Mechanical Fault Diagnosis: 2nd Edition)

► Show Figures

Figure 1

17 pages, 907 KB

Open AccessArticle

NeuroFusion-SLAM: A Deep Neural Network Framework for Real-Time Multi-Sensor SLAM

by Chenchen Yu, Wei Wei, Zhihong Cao, Zhiyuan Guo and Bo Fu

Sensors 2026, 26(7), 2267; https://doi.org/10.3390/s26072267 - 7 Apr 2026

Viewed by 828

Abstract

While deep learning-based visual SLAM (VSLAM) has achieved remarkable localization accuracy, its high computational cost and latency remain critical bottlenecks for real-time deployment. To address these limitations, this paper presents NeuroFusion-SLAM, a novel multi-sensor fusion framework tailored for both efficiency and robustness. By [...] Read more.

While deep learning-based visual SLAM (VSLAM) has achieved remarkable localization accuracy, its high computational cost and latency remain critical bottlenecks for real-time deployment. To address these limitations, this paper presents NeuroFusion-SLAM, a novel multi-sensor fusion framework tailored for both efficiency and robustness. By incorporating depthwise separable convolution, the framework cuts down model parameters by approximately 40% and training time by 49% while preserving localization accuracy, thus boosting real-time inference performance and computational efficiency in large-scale environments. Furthermore, a global edge optimization strategy is proposed by integrating sliding window optimization with a factor graph framework, which effectively improves the global consistency of the system. Extensive experiments on the TUM-VI and KITTI-360 datasets demonstrate that our system achieves real-time performance with an average latency of 30.4 ms per frame. It runs 3× faster than ORB-SLAM2 and 4× faster than VINS-Mono, while maintaining good localization accuracy. Full article

(This article belongs to the Section Navigation and Positioning)

► Show Figures

Figure 1

20 pages, 8292 KB

Open AccessArticle

Low-Complexity Noncoherent Demodulation Method for Underwater Electromagnetic Communication

by Longyang Deng, Deguang Zhao, Bizheng Liang, Xuhui Ding, Ziyi Yang and Dekang Liu

Sensors 2026, 26(7), 2266; https://doi.org/10.3390/s26072266 - 7 Apr 2026

Viewed by 707

Abstract

To strike a balance between complexity and performance in Minimum Shift Keying (MSK) systems for underwater electromagnetic communication, we propose a low-complexity maximum-likelihood (ML) noncoherent demodulation method. By integrating a resource reuse mechanism with a confidence-driven adaptive extension strategy, the proposed method significantly [...] Read more.

To strike a balance between complexity and performance in Minimum Shift Keying (MSK) systems for underwater electromagnetic communication, we propose a low-complexity maximum-likelihood (ML) noncoherent demodulation method. By integrating a resource reuse mechanism with a confidence-driven adaptive extension strategy, the proposed method significantly reduces computational resource consumption while maintaining near-optimal demodulation performance. Simulation results demonstrate that the bit-error-rate (BER) performance of the proposed method approaches that of the traditional fixed length ML receiver when the confidence threshold is set to 0.1. Meanwhile, the proposed method reduces complex correlation operations by 96.2% and complex addition operations by 87.1%, achieving minimal average computational overhead. Furthermore, we evaluate the method under frequency-flat Rayleigh fading channels, and the results confirm that the proposed method retains its performance advantage and complexity reduction under fading, supporting its potential for reliable underwater communication. Full article

(This article belongs to the Special Issue Recent Challenges in Underwater Optical Communication and Detection)

► Show Figures

Figure 1

25 pages, 5507 KB

Open AccessArticle

A Cheonjiin Layout Mental Speller: Developing a Simple and Cost-Effective EEG-Based Brain–Computer Interface System

by Ji Won Ahn, Gi Yeon Yu, Seong-Wan Kim, Young-Seek Seok, Kyung-Min Byun and Seung Ho Choi

Sensors 2026, 26(7), 2265; https://doi.org/10.3390/s26072265 - 7 Apr 2026

Cited by 1 | Viewed by 880

Abstract

A brain–computer interface (BCI) enables direct communication between the brain and external devices by translating neural activity into executable control commands. Among electroencephalography (EEG)-based paradigms, steady-state visual evoked potential (SSVEP) is widely adopted due to its high signal-to-noise ratio, robustness, and minimal calibration [...] Read more.

A brain–computer interface (BCI) enables direct communication between the brain and external devices by translating neural activity into executable control commands. Among electroencephalography (EEG)-based paradigms, steady-state visual evoked potential (SSVEP) is widely adopted due to its high signal-to-noise ratio, robustness, and minimal calibration requirements. While SSVEP-based spellers have been extensively investigated, many existing systems rely on high-channel-density EEG recordings and computationally complex processing pipelines, and are primarily designed for alphabetic input structures. In this study, we present an SSVEP-based Korean speller that integrates the Cheonjiin keyboard layout to support intuitive composition of Hangul syllables. The proposed system adopts a simple configuration, employing only five visual stimulation frequencies (6.67–12 Hz) and two occipital EEG channels (O1 and O2), with real-time frequency recognition performed using canonical correlation analysis (CCA) within a 1.5 s sliding window. EEG signals were acquired at 200 Hz using an OpenBCI Ganglion board, band-pass filtered (5–45 Hz), and processed with harmonic sinusoidal reference templates for multi-frequency classification. The proposed interface generates five control commands (up, down, left, right, and select), enabling directional cursor navigation and character confirmation on a 4 × 4 virtual Cheonjiin keyboard. Experimental validation with three healthy participants demonstrated an average classification accuracy of approximately 82% and an information transfer rate (ITR) of 31.2 bits/min. Frequency-domain analysis revealed clear spectral peaks at the stimulation frequencies and their harmonics, indicating reliable SSVEP responses. The proposed system employs a simple two-channel configuration integrated with a Korean language-specific input structure, demonstrating that reliable SSVEP-based communication can be realized without computationally intensive algorithms or high-cost EEG acquisition systems. These findings demonstrate that reliable SSVEP-based communication can be achieved using a low-channel configuration without reliance on high-cost EEG equipment. Full article

(This article belongs to the Section Electronic Sensors)

► Show Figures

Figure 1

20 pages, 4400 KB

Open AccessArticle

Tightly Coupled GNSS/IMU Hybrid Navigation Using Factor Graph Optimization with NLOS Detection Capability

by Haruki Tanimura and Toshiaki Tsujii

Sensors 2026, 26(7), 2264; https://doi.org/10.3390/s26072264 - 6 Apr 2026

Cited by 1 | Viewed by 958

Abstract

High-precision and reliable self-localization is essential for autonomous navigation systems. However, in urban canyons (urban environments with clusters of high-rise buildings), Global Navigation Satellite Systems (GNSS) suffer from severe multipath and Non-Line-of-Sight (NLOS) signal reception. This causes a theoretically unbounded positive bias in [...] Read more.

High-precision and reliable self-localization is essential for autonomous navigation systems. However, in urban canyons (urban environments with clusters of high-rise buildings), Global Navigation Satellite Systems (GNSS) suffer from severe multipath and Non-Line-of-Sight (NLOS) signal reception. This causes a theoretically unbounded positive bias in pseudorange measurements, significantly degrading positioning integrity. To address this challenge, this study proposes a novel GNSS/Inertial Measurement Unit (IMU) tightly coupled integrated navigation system using factor graph optimization (FGO) integrated with machine learning-based NLOS detection. To train the NLOS detection model, we utilized a dual-polarized antenna to label signals based on the strength difference between RHCP and LHCP components, achieving a detection accuracy of 0.89. A random forest classifier identifies NLOS signals, and based on its classification labels, the variance of the corresponding GNSS pseudorange factors within the FGO framework is dynamically inflated. This effectively mitigates the impact of outliers while preserving the graph topology. Experimental evaluations in dense urban environments demonstrated that the proposed method improves horizontal positioning accuracy by 84.8% compared to conventional standalone GNSS positioning. The dynamic integration of machine learning-based signal classification and tightly coupled FGO provides an extremely robust positioning solution, proven to meet the stringent reliability requirements demanded of autonomous systems even under severe signal obscuration. Full article

(This article belongs to the Special Issue Advances in GNSS/INS Integration for Navigation and Positioning)

► Show Figures

Figure 1

15 pages, 1901 KB

Open AccessArticle

DW-ReID: Vision–Language Learning for Person Re-Identification Under Diverse Weather Conditions

by Lei Cai, Yuying Liang, Bin Wang, Hexi Li, Jinquan Yang and Tao Zhu

Sensors 2026, 26(7), 2263; https://doi.org/10.3390/s26072263 - 6 Apr 2026

Viewed by 869

Abstract

Person re-identification (ReID) under diverse weather conditions remains a critical yet insufficiently explored problem. Most existing ReID approaches are developed and benchmarked on clear-weather datasets, resulting in significant performance degradation when deployed in rainy, snowy, or hazy environments. Conventional image restoration methods, typically [...] Read more.

Person re-identification (ReID) under diverse weather conditions remains a critical yet insufficiently explored problem. Most existing ReID approaches are developed and benchmarked on clear-weather datasets, resulting in significant performance degradation when deployed in rainy, snowy, or hazy environments. Conventional image restoration methods, typically optimized for low-level image quality metrics, are often misaligned with the objectives of high-level identity discrimination and thus fail to improve the person ReID performance. To address these limitations, we propose DW-ReID, a unified framework that integrates weather-degraded image restoration with person re-identification tasks. The proposed DW-ReID is built upon a large-scale Contrastive Language-Image Pre-training (CLIP) model and achieved by a two-stage training paradigm. In the first stage, a set of learnable text prompts is optimized to construct identity-specific ambiguous descriptions for each person’s identity. In the second stage, the optimized text descriptions, together with a frozen text encoder, provide language supervision to jointly train a weather encoder, an image restorer, and a ReID encoder in an end-to-end manner. The experimental results on two our contributed synthetic datasets consistently demonstrate the effectiveness and superior performance of the proposed DW-ReID method. Full article

(This article belongs to the Topic Applied Computer Vision and Pattern Recognition: 2nd Edition)

► Show Figures

Figure 1

19 pages, 4653 KB

Open AccessArticle

Nonlinear Ultrasonic Time-Domain Identification Based on Chaos Sensitivity and Its Application to Fatigue Detection of U71Mn Rail Steels

by Hongzhao Li, Mengfei Cheng, Chengzhong Luo, Weiwei Zhang, Jing Wu and Hongwei Ma

Sensors 2026, 26(7), 2262; https://doi.org/10.3390/s26072262 - 6 Apr 2026

Viewed by 590

Abstract

A nonlinear ultrasonic time-domain identification method based on chaos sensitivity was proposed in this study. The Duffing chaotic system was introduced into the weak second harmonic identification to realize early detection and quantitative evaluation of fatigue damage in U71Mn steel. First, to ensure [...] Read more.

A nonlinear ultrasonic time-domain identification method based on chaos sensitivity was proposed in this study. The Duffing chaotic system was introduced into the weak second harmonic identification to realize early detection and quantitative evaluation of fatigue damage in U71Mn steel. First, to ensure the reliability of nonlinear ultrasonic testing, a probe-pressure monitoring device was designed. Through pressure-stability experiments, 16 N was determined as the optimal pressure, which effectively suppresses contact nonlinearity interference and ensures coupling stability. Subsequently, the Duffing chaos detection system was established. The signal-system frequency-matching problem was resolved through time-scale transformation. Simultaneously, the issue of unknown initial phases was resolved using phase traversal compensation. Based on the chaotic system’s sensitivity to specific frequency signals and immunity to noise, the amplitudes of the fundamental wave and second harmonics in the target signals were quantified to calculate the nonlinear coefficient. Experimental results demonstrate that the proposed method can extract these amplitudes directly in the time domain, thereby effectively overcoming the spectral leakage inherent in traditional frequency-domain methods. The nonlinear coefficient of U71Mn steel exhibits a “double-peak” characteristic as fatigue damage increases. Specifically, the first peak appears at approximately 50% of fatigue life, while the second occurs at approximately 80%. This phenomenon is closely correlated with the distinct stages of internal fatigue crack propagation, reflecting a complex damage-evolution mechanism. This study not only provides a novel method for the precise extraction of weak nonlinear signals but also establishes a critical theoretical and experimental foundation for accurate fatigue life prediction for U71Mn rail steel. Full article

(This article belongs to the Section Fault Diagnosis & Sensors)

► Show Figures

Figure 1

18 pages, 2804 KB

Open AccessArticle

A Novel Contactless Scanning Conductivity-Detection Approach for Moving Reaction Boundary Analysis in Electrophoresis Titration Sensors

by Haozheng Dai, Youli Tian, Ke-Er Chen, Weiwen Liu, Qiang Zhang and Chengxi Cao

Sensors 2026, 26(7), 2261; https://doi.org/10.3390/s26072261 - 6 Apr 2026

Viewed by 643

Abstract

Electrophoresis titration sensors are widely used for biomarker detection. However, traditional methods rely on a visible boundary for signal readout. Although conventional capacitively coupled contactless conductivity detection avoids indicator dependence, its single-point detection method suffers from long measurement times, large amounts of redundant [...] Read more.

Electrophoresis titration sensors are widely used for biomarker detection. However, traditional methods rely on a visible boundary for signal readout. Although conventional capacitively coupled contactless conductivity detection avoids indicator dependence, its single-point detection method suffers from long measurement times, large amounts of redundant data, and the inability to dynamically monitor the moving reaction boundary. To address these issues, we developed a novel contactless scanning capacitively coupled conductivity-detection method for microchip electrophoretic titration sensors. This method enables the rapid tracking and monitoring of the boundary within the microfluidic channel through dynamic scanning. The spatial distribution of conductivity during electrophoretic titration was theoretically analyzed. To evaluate the method, glucose was chosen as a model analyte. Quantitative detection was achieved over the linear range of 0.2–50 mM, with the limit of detection of 0.1 mM. The method exhibited satisfactory stability with relative standard deviation values ranging from 0.9% to 4.3% (n = 3). While the detection limit is higher than optical methods (0.02 mM), the results confirmed that the novel method offers merits, such as compact size, low cost and label-free operation. Moreover, it demonstrated strong potential for portable, quantitative analysis of target analytes across a wide range of applications. Full article

(This article belongs to the Special Issue Sensors from Miniaturization of Analytical Instruments (3rd Edition))

► Show Figures

Figure 1

15 pages, 523 KB

Open AccessArticle

Artificial Neural Networks for Discrimination of Automotive Clear Coats by Vehicle Manufacturer

by Barry K. Lavine, Collin G. White and Douglas R. Heisterkamp

Sensors 2026, 26(7), 2260; https://doi.org/10.3390/s26072260 - 6 Apr 2026

Viewed by 752

Abstract

Modern automotive paints have a thin undercoat and color coat layer protected by a thick clear coat layer. All too often, only the clear coat layer of the automotive paint is recovered at the crime scene of a vehicle-related fatality. Searches for motor [...] Read more.

Modern automotive paints have a thin undercoat and color coat layer protected by a thick clear coat layer. All too often, only the clear coat layer of the automotive paint is recovered at the crime scene of a vehicle-related fatality. Searches for motor vehicle paint databases of clear coats using commercial software typically generate large hitlists that are difficult for a forensic paint examiner to work through unless additional information is provided for the search. To address this problem, deep learning has been applied to the infrared spectra of automotive clear coats to identify patterns in their spectra indicative of the motor vehicle manufacturer. An in-house automotive paint library of 2796 clear coat infrared spectra from six automotive manufacturers and 100 assembly plants was partitioned into training, validation, and prediction sets. Each spectrum has 1880 measurements over the spectral range of 4000 cm⁻¹ to 376 cm⁻¹. Several multilayer perceptron neural network models, each with three hidden layers, were developed that achieved high classification success rates for the training, validation, and prediction sets. The addition of convolutional layers to the deep learning neural network models did not improve the performance of these models. Full article

(This article belongs to the Special Issue Advanced Spectroscopy-Based Sensors and Spectral Analysis Technology)

► Show Figures

Graphical abstract

17 pages, 2174 KB

Open AccessArticle

RadarSSM: A Lightweight Spatiotemporal State Space Network for Efficient Radar-Based Human Activity Recognition

by Rubin Zhao, Fucheng Miao and Yuanjian Liu

Sensors 2026, 26(7), 2259; https://doi.org/10.3390/s26072259 - 6 Apr 2026

Viewed by 832

Abstract

Millimeter-wave radar has gradually gained popularity as a sensor mode for Human Activity Recognition (HAR) in recent years because it preserves the privacy of individuals and is resistant to environmental conditions. Nevertheless, the fast inference of high-dimensional and sparse 4D radar data is [...] Read more.

Millimeter-wave radar has gradually gained popularity as a sensor mode for Human Activity Recognition (HAR) in recent years because it preserves the privacy of individuals and is resistant to environmental conditions. Nevertheless, the fast inference of high-dimensional and sparse 4D radar data is still difficult to perform on low-resource edge devices. Current models, including 3D Convolutional Neural Networks and Transformer-based models, are frequently plagued by extensive parameter overhead or quadratic computational complexity, which restricts their applicability to edge applications. The present paper attempts to resolve these issues by introducing RadarSSM as a lightweight spatiotemporal hybrid network in the context of radar-based HAR. The explicit separation of spatial feature extraction and temporal dependency modeling helps RadarSSM decrease the overall complexity of computation significantly. Specifically, a spatial encoder based on depthwise separable 3D convolutions is designed to efficiently capture fine-grained geometric and motion features from voxelized radar data. For temporal modeling, a bidirectional State Space Model is introduced to capture long-range temporal dependencies with linear time complexity

O (T)

, thereby avoiding the quadratic cost associated with self-attention mechanisms. Extensive experiments conducted on public radar HAR datasets demonstrate that RadarSSM achieves accuracy competitive with state-of-the-art methods while substantially reducing parameter count and computational cost relative to representative convolutional baselines. These results validate the effectiveness of RadarSSM and highlight its suitability for efficient radar sensing on edge hardware. Full article

(This article belongs to the Special Issue Radar and Multimodal Sensing for Ambient Assisted Living)

► Show Figures

Figure 1

28 pages, 14521 KB

Open AccessArticle

Trajectory Prediction-Enabled Self-Decision-Making for Autonomous Cleaning Robots in Semi-Structured Dynamic Campus Environments

by Jie Peng, Zhengze Zhu, Qingsong Fan, Ranfei Xia and Zheng Yin

Sensors 2026, 26(7), 2258; https://doi.org/10.3390/s26072258 - 6 Apr 2026

Cited by 2 | Viewed by 1003

Abstract

Autonomous cleaning robots operating in semi-structured dynamic environments must execute task-oriented motions while safely interacting with surrounding agents. These agents include pedestrians, vehicles, and other robots. In such environments (e.g., interaction-rich campus environments), reliable self-decision-making requires anticipating the future motions of surrounding agents [...] Read more.

Autonomous cleaning robots operating in semi-structured dynamic environments must execute task-oriented motions while safely interacting with surrounding agents. These agents include pedestrians, vehicles, and other robots. In such environments (e.g., interaction-rich campus environments), reliable self-decision-making requires anticipating the future motions of surrounding agents rather than relying solely on reactive obstacle avoidance. This paper presents a trajectory prediction-enabled self-decision-making framework for autonomous cleaning robots in campus environments. A learning-based multi-agent trajectory prediction model is trained offline using public benchmarks and real-world operational data to capture typical interaction patterns in corridor-following, edge-cleaning, and intersection scenarios. The predicted trajectories are then incorporated as forward-looking priors into the robot’s online decision-making and planning process, enabling prediction-aware yielding, detouring, and task continuation decisions. The proposed framework is evaluated using real-world data-driven scenario reconstruction on a high-fidelity simulation platform that incorporates realistic vehicle dynamics and heterogeneous traffic participants. This evaluation focuses on short-horizon prediction performance and its impact on downstream decision-making stability. The results show that integrating trajectory prediction into the decision-making loop leads to more stable motion behavior and fewer abrupt adjustments in interaction scenarios. Under short-term prediction horizons, the evaluation results show that the proposed model achieves ADERate and FDERate exceeding 90% under predefined error thresholds, while lane-change prediction accuracy remains around 79%. In addition, the robot maintains stable speed tracking with only minor fluctuations under medium-density traffic conditions. Full article

(This article belongs to the Special Issue Robot Swarm Collaboration in the Unstructured Environment)

► Show Figures

Figure 1

29 pages, 10248 KB

Open AccessArticle

Fs2PA: A Full-Scale Feature Synergistic Perception Architecture for Vehicular Infrared Object Detection via Physical Priors and Semantic Constraints

by Boxuan Pei, Leyuan Wu, Xiaoyan Zheng, Chao Zhou and Dingxiang Wang

Sensors 2026, 26(7), 2257; https://doi.org/10.3390/s26072257 - 6 Apr 2026

Cited by 1 | Viewed by 564

Abstract

Vehicular infrared object detection is a key technology supporting autonomous driving systems to achieve all-weather environmental perception. However, infrared images inherently lack texture, resulting in blurred object contours. Additionally, deep network propagation severely erodes and loses feature information of distant tiny objects. To [...] Read more.

Vehicular infrared object detection is a key technology supporting autonomous driving systems to achieve all-weather environmental perception. However, infrared images inherently lack texture, resulting in blurred object contours. Additionally, deep network propagation severely erodes and loses feature information of distant tiny objects. To address the above issues, this study proposes a Full-Scale Feature Synergistic Perception Architecture for vehicular infrared object detection. This architecture first designs a Gradient-Informed Attention module, which initializes convolution kernels through physical gradient operators to inject geometric prior information into the network, enhancing the model’s perception capability of blurred object boundaries. Secondly, it constructs a Full-Scale Feature Pyramid containing a

P_{2}

high-resolution feature layer to effectively recover the geometric detail features of distant tiny objects. Finally, it proposes a Scale-Aware Shared Head, which relies on a cross-scale parameter sharing mechanism to achieve extreme parameter compression, and simultaneously introduces deep semantic information to form strong constraints, suppressing noise interference in shallow features. Experimental results on the FLIR v2 and M3FD datasets show that the proposed architecture exhibits excellent detection performance. On FLIR v2, it raises mAP@50 to 64.06% (6.51% relative gain vs. YOLOv11) while maintaining 547 FPS inference speed, achieving an optimal accuracy–efficiency balance. Full article

(This article belongs to the Special Issue AI Agent Driven Sensing, Data Acquisition, and Signal Processing Methods in Autonomous Driving)

► Show Figures

Graphical abstract

21 pages, 5751 KB

Open AccessArticle

A Hybrid VMD-Transformer-BiLSTM Framework with Cross-Attention Fusion for Aileron Fault Diagnosis in UAVs

by Yang Song, Weihang Zheng, Xiaoyu Zhang and Rong Guo

Sensors 2026, 26(7), 2256; https://doi.org/10.3390/s26072256 - 6 Apr 2026

Viewed by 756

Abstract

Aileron fault diagnosis in fixed-wing unmanned aerial vehicles (UAVs) faces significant challenges due to strong noise, multi-modal coupling, and limited fault samples. This paper presents a hybrid fault diagnosis framework that integrates variational mode decomposition (VMD) with a cross-attention-based feature fusion mechanism. First, [...] Read more.

Aileron fault diagnosis in fixed-wing unmanned aerial vehicles (UAVs) faces significant challenges due to strong noise, multi-modal coupling, and limited fault samples. This paper presents a hybrid fault diagnosis framework that integrates variational mode decomposition (VMD) with a cross-attention-based feature fusion mechanism. First, residual signals are generated from UAV kinematic models and decomposed into multi-scale intrinsic mode functions (IMFs) using VMD to extract multiscale frequency-localized features. An integrated framework is then constructed, where Transformer encoders capture the global features and bidirectional long short-term memory (BiLSTM) networks extract local temporal dynamics. To effectively combine these complementary features, a cross-attention fusion module is designed to focus on the discriminative time-frequency features. Furthermore, a hybrid pooling strategy integrating max pooling and attention pooling is introduced to enhance classification robustness. Experiments on the AirLab failure and anomaly (ALFA) dataset demonstrate that the proposed method achieves 95.12% accuracy with improved fault separability, outperforming VMD + BiLSTM (87.66%), VMD + Transformer (86.89%), Transformer + BiLSTM (84.83%), Transformer (72.24%), CNN + LSTM (94.05%), and HDMTL (94.86%). Full article

(This article belongs to the Section Fault Diagnosis & Sensors)

► Show Figures

Figure 1

19 pages, 919 KB

Open AccessArticle

A Sequential Kalman-Newton-KM Framework for AIS and Radar Data Fusion in Restricted Inland Waterways

by Huixia Shi, Dejun Wang, Longting Wei and Shan Liang

Sensors 2026, 26(7), 2255; https://doi.org/10.3390/s26072255 - 6 Apr 2026

Cited by 1 | Viewed by 809

Abstract

This paper presents a novel data fusion framework that integrates Automatic Identification System (AIS) data with radar surveillance for real-time vessel monitoring in inland restricted waterways. The approach exploits the complementarity between heterogeneous sensors: AIS provides semantic information with temporal sparsity, while radar [...] Read more.

This paper presents a novel data fusion framework that integrates Automatic Identification System (AIS) data with radar surveillance for real-time vessel monitoring in inland restricted waterways. The approach exploits the complementarity between heterogeneous sensors: AIS provides semantic information with temporal sparsity, while radar offers high-frequency observations without vessel identity. The proposed solution combines Kalman filtering and Newton interpolation (K-N) for high-resolution AIS resampling, followed by optimal data association using the Kuhn-Munkres (KM) algorithm. By formulating data association as a global optimization problem, the framework achieves globally optimal sensor fusion while effectively handling data imbalance through virtual point augmentation. Experimental validation using real-world data demonstrates a matching accuracy of 94.2% in low-density scenarios and 80.1% in high-traffic conditions, with computational efficiency suitable for real-time deployment. The system performs consistently across different waterway geometries, although performance varies slightly between curved and straight channels. By fusing the high temporal resolution of radar data with the rich identity information from AIS, this framework enables more accurate and reliable vessel tracking, providing waterway authorities with enhanced situational awareness for improved traffic management and scheduling in restricted waterways. Full article

(This article belongs to the Special Issue Intelligent Transportation Systems with Connected Vehicle, Cloud Computing, and Internet of Things Technologies)

► Show Figures

Figure 1

35 pages, 4925 KB

Open AccessArticle

Defect-Mask2Former: An Improved Semantic Segmentation Model for Precise Small-Sized Defect Detection on Large-Sized Timbers

by Mingming Qin, Hongxu Li, Yuxiang Huang, Xingyu Tong and Zhihong Liang

Sensors 2026, 26(7), 2254; https://doi.org/10.3390/s26072254 - 6 Apr 2026

Cited by 1 | Viewed by 1072

Abstract

The precise segmentation of small-sized defects on wood surfaces is critical for the quality grading of glued laminated timber (GLT). Existing semantic segmentation models face core bottlenecks in this context: high miss rates, blurred boundary localization, and excessive size measurement errors. To address [...] Read more.

The precise segmentation of small-sized defects on wood surfaces is critical for the quality grading of glued laminated timber (GLT). Existing semantic segmentation models face core bottlenecks in this context: high miss rates, blurred boundary localization, and excessive size measurement errors. To address these issues, this paper proposes an improved Defect-Mask2Former model that integrates an Attention-Guided Pyramid Enhancement (AGPE) module and a Defect Boundary Calibration and Correction (DBCC) module. Through synergistic optimization, the model achieved pixel-level precise segmentation. To support model training and validation, a custom image acquisition device was designed, and the PlankDefSeg dataset was constructed, comprising 3500 pixel-level annotated images covering five defect types across six industrial wood species. Experimental results demonstrate that on the PlankDefSeg dataset, Defect-Mask2Former achieved a mean Intersection over Union (mIoU) of 85.34% for small-sized defects, a 17.84% improvement over the baseline Mask2Former. The miss rate was reduced from 20.78% to 5.83%, and the size measurement error was only 2.86%, strictly meeting the ≤3% accuracy requirement of the GB/T26899-2022 standard. The model achieved an inference speed of 27.6 FPS, satisfying real-time detection needs. By integrating the model into the GLT grading workflow, a grading accuracy of 94.3% was achieved, and the processing time per timber was reduced from 30 s to 1.5 s, a 20-fold efficiency improvement. This study provides reliable technical support for intelligent GLT quality grading and offers a reference solution for other industrial surface defect segmentation tasks. Full article

(This article belongs to the Section Smart Agriculture)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Sensors, Volume 26, Issue 7 (April-1 2026) – 273 articles

Further Information

Guidelines

MDPI Initiatives

Follow MDPI