ieeexplore最新文献

筛选
英文 中文
Burial Depth Estimation for Partially Embedded Rocks Using Scanning Laser Doppler Vibrometry and Neural Networks 基于扫描激光多普勒振动和神经网络的部分嵌套岩石埋深估计
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-18 DOI: 10.1109/JSEN.2026.3733381
Yiyan Ruan;Erik Komendera
{"title":"Burial Depth Estimation for Partially Embedded Rocks Using Scanning Laser Doppler Vibrometry and Neural Networks","authors":"Yiyan Ruan;Erik Komendera","doi":"10.1109/JSEN.2026.3733381","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3733381","url":null,"abstract":"Estimating the burial depth of partially embedded rocks is a key capability for energy-limited robotic excavation, where unnecessary digging or repeated probing can be prohibitively costly. This article presents a geometry-informed, vibration-based pipeline that combines impulse excitation, scanning laser Doppler vibrometry (LDV), and exposed geometry to estimate burial depth with a neural network. Unlike spring–mass inversions, which require rock mass and soil stiffness, the proposed estimator uses no measured mass, assumed density, or independently measured soil parameters as model inputs. A multilayer perceptron (MLP) maps 11 measurable features, including resonance parameters, half-power damping ratio, and spatial vibration decay, to burial depth. Under strict leave-one-rock-out (LORO) cross-validation with nested hyperparameter selection, it achieves 25.1% mean absolute percentage error (MAPE) and 0.94 cm root-mean-squared error (RMSE) on 18 near-uniform-density (1.92–2.03 g/cm<sup>3</sup>) concrete blocks spanning 204 measurements in one controlled sand condition. The mean signed error is <inline-formula> <tex-math>$+{0}.{01}$ </tex-math></inline-formula> cm, the absolute error is roughly 1 cm across the 1–8.4 cm depth range, and the largest observed error across the 20 seeded nested cross-validation runs is 3.0 cm. The estimator also outperforms a calibrated closed-form inversion supplied with the true rock mass (45.9% MAPE). To the authors’ knowledge, this is the first mass-agnostic, data-driven burial-depth estimator evaluated under cross-specimen LORO validation.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"33392-33402"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315514","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Multi-Covariate Imaging of Sub-Resolution Targets Combined With Minimum Variance-Based Adaptive Weighting for Ultrafast Ultrasound Imaging 基于最小方差自适应加权的亚分辨率目标多协变量成像超快超声成像
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-24 DOI: 10.1109/JSEN.2026.3735510
Yuanguo Wang;Pei Tang;Yadan Wang;Zhihui Han;Chichao Zheng;Hu Peng
{"title":"Multi-Covariate Imaging of Sub-Resolution Targets Combined With Minimum Variance-Based Adaptive Weighting for Ultrafast Ultrasound Imaging","authors":"Yuanguo Wang;Pei Tang;Yadan Wang;Zhihui Han;Chichao Zheng;Hu Peng","doi":"10.1109/JSEN.2026.3735510","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3735510","url":null,"abstract":"The multi-covariate imaging of sub-resolution targets (MIST) has demonstrated consistent enhancements in terms of contrast-to-noise ratio (CNR) and speckle statistics in ultrasound imaging. However, the resolution of MIST is approximately equivalent or slightly worse than that of the conventional delay-and-sum (DAS), and the contrast improvement of MIST is insufficient at larger depths. To address these issues, we propose two novel adaptive weighting approaches for MIST in plane-wave (PW) ultrasound imaging. In the proposed method, a dynamic subarray length determined by amplitude standard deviation and coherence is used for covariance matrix estimation in eigenspace-based minimum variance (EIBMV), and the eigenvector associated with the largest eigenvalue is used to reconstruct the weight vector. Then, pixel-based adaptive EIBMV weight (AEW) and convolutional AEW (CAEW) are formed by summing the weight vector and averaging the convolutional weight vector, respectively. We validate the proposed methods using simulated, experimental phantom, and in vivo datasets. Compared with MIST, the proposed CAEW-MIST achieves average improvements of 51.2%, 67.2%, 1.0%, and 18.0% in resolution, contrast, generalized CNR (gCNR), and speckle signal-to-noise ratio (sSNR), respectively, in phantom experiments. These findings demonstrate that the proposed methods can significantly enhance the overall imaging quality of MIST in ultrafast ultrasound imaging.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"33161-33175"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315573","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Sensing Information-Assisted CSI Feedback for UAV Communication Systems 无人机通信系统传感信息辅助CSI反馈
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2025-09-15 DOI: 10.1109/JSEN.2025.3607326
Chaojin Qing;Yu Sun;Yuqiao Yang;Haowen Jiang;Pengfei Du;Mian Ye
{"title":"Sensing Information-Assisted CSI Feedback for UAV Communication Systems","authors":"Chaojin Qing;Yu Sun;Yuqiao Yang;Haowen Jiang;Pengfei Du;Mian Ye","doi":"10.1109/JSEN.2025.3607326","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3607326","url":null,"abstract":"In uncrewed aerial vehicle (UAV)-assisted frequency-division duplex (FDD) wireless communication systems, accurate channel state information (CSI) is critical to harnessing the benefits of massive multiple-input–multiple-output (mMIMO). However, CSI feedback in UAV-assisted FDD mMIMO systems still faces significant challenges, including high feedback overhead, inevitable channel estimation (CE) errors at the UAV, and complex reconstruction algorithms at the ground base station (gBS). To address these challenges, this article proposes a dual-branch CSI feedback method assisted by sensing information, inspired by integrated sensing and communication (ISAC). First, the UAV’s location and speed are extracted as prior sensing information from the global navigation satellite system (GNSS). Based on this information, we develop a sensing-aided two-stage CE error suppression method in the delay-Doppler (DD) domain, which incorporates initial and threshold-based error suppression techniques. Notably, a dual-branch compression scheme is designed to significantly reduce feedback overhead by leveraging the sensed prior information. At the gBS, a low-complexity CSI recovery algorithm is proposed. Benefiting from the dual-branch compression scheme, downlink CSI can be recovered through simple mapping operations or a few iterations, thereby reducing computational complexity while improving recovery accuracy. Experimental results demonstrate that the proposed method effectively enhances normalized mean-square-error (NMSE) performance while reducing both computational complexity and feedback overhead. The method also exhibits robustness against parameter variations. Additionally, it is compatible with deep learning (DL)-based CSI feedback frameworks.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"31703-31715"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A GaN/Graphene-Oxide Composite QBIC Terahertz Metasensor for Enhanced Trace Small-Molecule Detection 氮化镓/氧化石墨烯复合QBIC太赫兹超传感器增强痕量小分子检测
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-23 DOI: 10.1109/JSEN.2026.3734940
Yan Wang;Qike Shi;Shide Zhang;Xin Jue;Lanju Liang;Haiyun Yao;Xin Yan;Guifang Wu;Zhenhua Li;Xiaofei Hu;Ziyuan Gao;Lei Bao
{"title":"A GaN/Graphene-Oxide Composite QBIC Terahertz Metasensor for Enhanced Trace Small-Molecule Detection","authors":"Yan Wang;Qike Shi;Shide Zhang;Xin Jue;Lanju Liang;Haiyun Yao;Xin Yan;Guifang Wu;Zhenhua Li;Xiaofei Hu;Ziyuan Gao;Lei Bao","doi":"10.1109/JSEN.2026.3734940","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3734940","url":null,"abstract":"Terahertz (THz) metasurfaces (MSs) provide a promising platform for highly sensitive and label-free biosensing. However, the weak THz response and limited surface loading of low-molecular-weight analytes hinder trace detection. Here, L-phenylalanine (L-Phe) is used as a representative small molecule to evaluate an ultrasensitive THz biosensing platform based on enhanced molecular enrichment and weak-perturbation amplification. Although quasibound states in the continuum (QBIC) MS biosensors provide pronounced resonant field confinement and enhanced light–matter interactions, the weak perturbation of the local dielectric environment induced by limited molecular adsorption makes the trace detection of small-molecule analytes challenging. To address this critical issue, this article proposes a composite structure MS biosensor based on the QBIC resonance. The sensor innovatively integrates a patterned gallium nitride (GaN) and graphene oxide (GO) bifunctional layer to overcome the inherent performance limitations of existing biosensors in the detection of small biomolecules. The GO layer enriches L-Phe molecules through <inline-formula> <tex-math>$pi $ </tex-math></inline-formula>–<inline-formula> <tex-math>$pi $ </tex-math></inline-formula> interactions, hydrogen bonding, and possible electrostatic interactions, while the patterned GaN layer provides localized dielectric loading and serves as a supporting interface for GO integration. Molecular adsorption modifies the effective permittivity and dielectric loss of the local GaN-GO sensing region. These weak dielectric perturbations are amplified by the strong near-field confinement of the QBIC mode, producing pronounced THz transmission responses. The integrated sensor achieves a maximum modulation depth of 37%, approximately 3.7 times that of the bare metallic MSs, with 3.6 fg mL<sup>−1</sup> representing the lowest experimentally tested concentration. These results demonstrate the synergistic effects of GaN dielectric loading, GO-assisted molecular enrichment, and QBIC field confinement.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"32379-32387"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Integrated Microwave Filter–Sensor With Dual-Frequency Consistency Verification for High-Sensitivity Solid-Material Permittivity Characterization 具有双频一致性验证的集成微波滤波传感器用于高灵敏度固体材料介电常数表征
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-29 DOI: 10.1109/JSEN.2026.3736803
Dian Rusdiyanto;Zahriladha Zakaria;Gunawan Wibisono
{"title":"Integrated Microwave Filter–Sensor With Dual-Frequency Consistency Verification for High-Sensitivity Solid-Material Permittivity Characterization","authors":"Dian Rusdiyanto;Zahriladha Zakaria;Gunawan Wibisono","doi":"10.1109/JSEN.2026.3736803","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3736803","url":null,"abstract":"Existing integrated microwave filter–sensor designs commonly rely on a single sensing resonance or locate the sensing response close to the filtering passband, which can cause response ambiguity and residual interaction during material loading. This work presents a compact two-port integrated microwave filter–sensor with dual-frequency consistency verification for solid-material permittivity characterization. The proposed structure combines a folded stepped-impedance resonator (SIR) that produces two bandpass filter (BPF) passbands at 2.1 and 2.45 GHz with two split-ring-resonator (SRR)-based sensing resonators operating at 1.5 and 3.5 GHz. By positioning the sensing resonances outside the BPF passbands, the design preserves the filtering responses while enabling separate lower band and upper band material sensing. Equivalent-circuit modeling and electric-field distribution analysis verify that the filtering and sensing responses originate from different physical regions. The lower band microwave sensor (LB-MS) and upper band microwave sensor (UB-MS) achieve maximum frequency detection resolutions (FDRs) of 72 MHz/<inline-formula> <tex-math>$boldsymbol {varepsilon }_{mathbf {r}}$ </tex-math></inline-formula> and 125 MHz/<inline-formula> <tex-math>$boldsymbol {varepsilon }_{mathbf {r}}$ </tex-math></inline-formula>, with normalized sensitivities of 4.87% and 3.59%, respectively. Measurements using RO5880, RO4835, FR4, and TMM10i show permittivity-extraction accuracies of 92.21%–98.82% for LB-MS and 91.20%–97.00% for UB-MS. The dual-frequency consistency deviation, <inline-formula> <tex-math>$boldsymbol {D}_{mathbf {FC}}$ </tex-math></inline-formula>, ranges from 1.09% to 3.36% for the intact samples and reaches a maximum of 4.19% among the investigated defected FR4 samples, demonstrating integrated RF filtering, dual-frequency sensing, consistency-verified permittivity extraction, and evaluation of defect-induced changes in extracted permittivity in one compact microwave platform.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"32567-32576"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Learning-Based Full-Field Shape Sensing for Flexible Structures Using Orthogonal FBG Arrays 基于正交FBG阵列的柔性结构学习全场形状传感
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-24 DOI: 10.1109/JSEN.2026.3735339
Wen Wang;Xinhuai Wang;Jiayu Chen;Yiteng Li;Junxian Chen;Zhizhou Long;Yin Xu;Tian Ma
{"title":"Learning-Based Full-Field Shape Sensing for Flexible Structures Using Orthogonal FBG Arrays","authors":"Wen Wang;Xinhuai Wang;Jiayu Chen;Yiteng Li;Junxian Chen;Zhizhou Long;Yin Xu;Tian Ma","doi":"10.1109/JSEN.2026.3735339","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3735339","url":null,"abstract":"Flexible structures undergo free-form deformation that is difficult to measure densely during operation. This article presents learning-based full-field shape sensing with an orthogonal fiber Bragg grating (FBG) array and a residual convolutional encoder–decoder. Paired FBG and depth measurements provide offline supervision, and deployment uses FBG measurements alone. Evaluation applies training-partition normalization under a frame-level benchmark and a separately retrained complete-acquisition-file holdout protocol. At 128-by-128 output resolution, the frame-level benchmark gives mean and median root-mean-square errors (RMSEs) of 1.79 and 1.10 mm; 41 held-out acquisition files give an equal-file mean RMSE of 4.49 mm and a mean structural similarity index (SSIM) of 0.856. Matched baselines support the residual model, and ablation identifies the benefit of residual refinement. Software-masked input experiments show that accuracy is sensitive to retained count, spatial layout, and directional allocation. A selected compact subset yields a frame-level RMSE point estimate close to the full-input result, while equal-count layouts yield different errors and single-direction performance varies with acquisition condition. The selected graphics processing unit (GPU) path requires 4.37 ms per prediction, and recorded-sequence replay characterizes one changed acquisition regime. These results establish a sparse-to-dense reconstruction route for flexible structures within the tested topology and acquisition conditions.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"32388-32396"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Movable-Antenna-Assisted Cooperative ISAC in Near-Field Scenarios: Joint Transmit Beamforming and Antenna Positioning 近场场景下移动天线辅助的协同ISAC:联合发射波束形成和天线定位
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-02-09 DOI: 10.1109/JSEN.2026.3660782
Xinru Wang;Xiaobo Liu;Fei Du;Suiyan Geng;Zhenyu Zhou
{"title":"Movable-Antenna-Assisted Cooperative ISAC in Near-Field Scenarios: Joint Transmit Beamforming and Antenna Positioning","authors":"Xinru Wang;Xiaobo Liu;Fei Du;Suiyan Geng;Zhenyu Zhou","doi":"10.1109/JSEN.2026.3660782","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3660782","url":null,"abstract":"This work investigates a multiple base station (BS) cooperative integrated sensing and communication (ISAC) system assisted by movable antennas (MAs) to enhance spatial degrees of freedom (DoFs). Given the larger antenna apertures introduced by MAs, the traditional far-field channel assumption between transceivers becomes invalid, and thus, the channel model is extended to near-field propagation incorporating distance-dependent phase variations. To guarantee cooperative sensing performance, we formulate a robust optimization problem to jointly design multi-BS transmit beamforming and antenna positioning to maximize the worst case sensing signal-to-noise ratio (SNR), accounting for both target location uncertainty and channel state information (CSI) estimation error. To tackle this nonconvex problem, we first reformulate it using convex hull approximation for location uncertainty and the S-Procedure for CSI error, transforming constraints into tractable forms. We then propose an efficient alternating optimization (AO) framework that decouples the variables into subproblems, which includes updating worst case weights via closed-form expressions derived from the reverse Hölder inequality, solving the beamforming subproblem via the sequential rank-one constraint relaxation (SROCR) method, and optimizing the antenna positioning subproblem using the minorization–maximization (MM) method to construct surrogates for the nonconvex elements. Numerical results validate the effectiveness of the proposed algorithm and demonstrate that the proposed scheme achieves significant improvements in sensing performance compared to noncooperative baselines. Furthermore, by exploiting enhanced spatial DoFs through antenna repositioning, the proposed scheme also yields substantial gains over fixed-position antenna (FPA) baselines.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"31849-31863"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Wearable Sensor-Based Human Activity Recognition in Squash Training Optimized by Reinforcement Learning 基于可穿戴传感器的强化学习优化壁球训练人体活动识别
IF 4.5 2区 综合性期刊
IEEE Sensors Journal Pub Date : 2026-11-01 Epub Date: 2026-09-11 DOI: 10.1109/JSEN.2026.3730906
Haodong Huang;Chengshang Si;Shilong Sun;Yufan Wang;Xiao Zhang;Yu Zhou;Wenzhao Zhang;Dong Wang
{"title":"Wearable Sensor-Based Human Activity Recognition in Squash Training Optimized by Reinforcement Learning","authors":"Haodong Huang;Chengshang Si;Shilong Sun;Yufan Wang;Xiao Zhang;Yu Zhou;Wenzhao Zhang;Dong Wang","doi":"10.1109/JSEN.2026.3730906","DOIUrl":"https://doi.org/10.1109/JSEN.2026.3730906","url":null,"abstract":"This study presents a measurement and classification system for human activity recognition (HAR) in squash training, integrating inertial sensor data with a Q-learning-based learning-rate optimization strategy for accurate squash action recognition. In this approach, an inertial measurement unit (IMU) is attached to the wrist to acquire real-time kinematic information during squash movements. A dual-loop Kalman filter is employed to preprocess the sensor signals and quantify signal confidence using a statistical elliptical threshold, enabling reliable segmentation of individual action instances while reducing noise-induced uncertainty. Subsequently, statistical features are extracted to construct robust motion representations. Based on these representations, a temporal shift convolutional neural network (TS-CNN) optimized by the proposed Q-learning-based learning-rate selection strategy is developed for squash action classification. Experimental results demonstrate that the proposed system achieves an average classification accuracy of 96.94% over five independent runs, outperforming conventional machine learning and deep learning-based approaches. Furthermore, the robustness of the proposed measurement pipeline under varying experimental conditions is validated. This study provides an effective solution for wearable sensor-based squash motion recognition and contributes to the development of intelligent sports analytics.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"26 21","pages":"32723-32737"},"PeriodicalIF":4.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149315895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
SCARF: Stage-Wise Clarification-Based Ambiguity-Resolving Framework for NL Control of Semiconductor Packaging Equipment 基于阶段澄清的半导体封装设备NL控制的歧义解决框架
IF 2.5 3区 工程技术
IEEE Transactions on Semiconductor Manufacturing Pub Date : 2026-11-01 Epub Date: 2026-09-14 DOI: 10.1109/TSM.2026.3733702
Heeyeon Joo;Jaewook Lee;Donghee Han;Mujin Kim;Keejun Han
{"title":"SCARF: Stage-Wise Clarification-Based Ambiguity-Resolving Framework for NL Control of Semiconductor Packaging Equipment","authors":"Heeyeon Joo;Jaewook Lee;Donghee Han;Mujin Kim;Keejun Han","doi":"10.1109/TSM.2026.3733702","DOIUrl":"https://doi.org/10.1109/TSM.2026.3733702","url":null,"abstract":"Modern semiconductor packaging inspection equipment (SPIE) exposes hundreds of machine-control endpoints governing inspection recipes, creating a complex control burden for operators of varying experience levels. While large language models (LLMs) have been applied to natural-language equipment control, they exhibit critical failure modes under shop-floor conditions: domain-specific parameter hallucination, unverified dispatch of under-specified commands, and incompatibility with air-gapped network environments. We propose SCARF (Stage-wise Clarification-based Ambiguity-Resolving Framework), a three-stage on-device pipeline that performs concept-level intent retrieval, margin-based ambiguity classification, and schema-constrained slot filling, and constrained command construction that combines grammar-constrained endpoint selection with schema-constrained slot filling, dispatching a control command to the equipment only after it passes endpoint- and value-level validation, and issuing a targeted clarification question otherwise. On a 300-utterance benchmark collected from a deployed GVision inspection tool (a vision-based SPIE inspection UI), SCARF achieves 76.33% end-to-end URL exact-match accuracy and reduces the unsafe dispatch rate on ambiguous commands to 3.00%—the lowest among all evaluated systems, including cloud-hosted LLM baselines—at a mean latency of 0.74 s on commodity CPU hardware, a <inline-formula> <tex-math>$58times $ </tex-math></inline-formula> reduction over direct LLM inference and comparable in accuracy to prompting pipelines that require cloud GPU resources. These results demonstrate that clarification-gated, offline hybrid inference is a viable and controllable operating point for safety-critical natural-language equipment control in high-volume manufacturing environments.","PeriodicalId":451,"journal":{"name":"IEEE Transactions on Semiconductor Manufacturing","volume":"39 4","pages":"886-903"},"PeriodicalIF":2.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149399510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A Physics-Constrained, Design-Driven Methodology for Defect Dataset Generation in Optical Lithography 光学光刻中缺陷数据集生成的物理约束、设计驱动方法
IF 2.5 3区 工程技术
IEEE Transactions on Semiconductor Manufacturing Pub Date : 2026-11-01 Epub Date: 2026-09-03 DOI: 10.1109/TSM.2026.3729945
Yuehua Hu;Jiyeong Kong;Pangyun Jeong;Dohee Jeong;Dong-Yeol Shin;Jaekyun Kim;Kyung-Tae Kang
{"title":"A Physics-Constrained, Design-Driven Methodology for Defect Dataset Generation in Optical Lithography","authors":"Yuehua Hu;Jiyeong Kong;Pangyun Jeong;Dohee Jeong;Dong-Yeol Shin;Jaekyun Kim;Kyung-Tae Kang","doi":"10.1109/TSM.2026.3729945","DOIUrl":"https://doi.org/10.1109/TSM.2026.3729945","url":null,"abstract":"Artificial intelligence (AI)-based defect inspection in micro/nano manufacturing is limited by the scarcity of high quality, physically grounded training data. To address this bottleneck, this study proposes a design-driven and fabrication-based methodology that generates lithographic defects through mathematically defined morphology operations and physically replicates them using digital micromirror device (DMD)-based maskless lithography. Bridge, burr, and pinch defects were systematically generated from defect-free layouts using mathematical morphology and physically replicated using DMD-based maskless lithography. The fabricated defect patterns were compared with their defect-free references to obtain pixel-level annotations. The resulting dataset contains 3,530 optical micrographs and 13,404 annotated defect instances, including morphology-generated bridge, burr, and pinch defects and naturally occurring contamination. Quantitative analysis of 3,000 matched defect pairs showed positive correspondence between designed and fabricated defect sizes, with a Pearson correlation coefficient of 0.760 and a Spearman rank correlation coefficient of 0.728. Five instance segmentation models were evaluated using a base-layout-disjoint split, achieving AP@0.50 values of 0.704 to 0.990 across the morphology-generated defect classes, while contamination remained more challenging. Qualitative evaluation on experimentally fabricated ICCAD-derived layouts showed that models trained on the constructed dataset could localize defect regions in previously unseen and more complex patterns. These results demonstrate that mathematically controlled design perturbations are measurably reflected in fabricated patterns and support the proposed methodology as a physically grounded approach to lithography inspection dataset construction.","PeriodicalId":451,"journal":{"name":"IEEE Transactions on Semiconductor Manufacturing","volume":"39 4","pages":"822-831"},"PeriodicalIF":2.5,"publicationDate":"2026-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149399522","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
相关产品
×
本文献相关产品
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信
小红书