<rss version="2.0" xmlns:atom="https://www.w3.org/2005/Atom">
  <channel>
    <title>Transport Research International Documentation (TRID)</title>
    <link>https://trid.trb.org/</link>
    <atom:link href="https://trid.trb.org/Record/RSS?s=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" rel="self" type="application/rss+xml" />
    <description></description>
    <language>en-us</language>
    <copyright>Copyright © 2026. National Academy of Sciences. All rights reserved.</copyright>
    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
    <managingEditor>tris-trb@nas.edu (Bill McLeod)</managingEditor>
    <webMaster>tris-trb@nas.edu (Bill McLeod)</webMaster>
    <image>
      <title>Transport Research International Documentation (TRID)</title>
      <url>https://trid.trb.org/Images/PageHeader-wTitle.jpg</url>
      <link>https://trid.trb.org/</link>
    </image>
    <item>
      <title>A traffic incident detection method of highway based on spatial-temporal characteristics</title>
      <link>https://trid.trb.org/View/2701207</link>
      <description><![CDATA[In this study, we propose a method for detecting traffic incidents by leveraging tensor decomposition in conjunction with spatial-temporal constraints. The method comprises three main stages. Firstly, we develop an initialization and noise reduction procedure for the traffic data pre-processing model, which includes Tucker decomposition and truncating higher-order Singular Value Decomposition (SVD). Subsequently, we construct spatial and temporal coefficient matrices based on the Pearson test and introduce a Laplace penalty term to quantify the disparity between predicted and actual data. Next, we validate the method using traffic loop detector data and incident records from I90 in Seattle, USA, in 2015, comparing its performance with seven other traffic prediction methods. The results demonstrate the superior prediction accuracy, correct detection rate, and low false alarm rate of our traffic incident detection method. Specifically, the mean square error of speed prediction is 5.22km/h, the average relative error is 6.88%, and the detection rate reaches 98.92%. Our method effectively evaluates the spatial-temporal characteristics of traffic data, enabling accurate prediction and detection of traffic incidents. Its technical applicability holds promise for enhancing the capacity and efficiency of future traffic control systems.]]></description>
      <pubDate>Wed, 17 Jun 2026 16:14:27 GMT</pubDate>
      <guid>https://trid.trb.org/View/2701207</guid>
    </item>
    <item>
      <title>Enhanced Sensorless Position Estimation for Dual Three-Phase PMSM via Terminal Voltage Detection-Based High-Frequency Injection</title>
      <link>https://trid.trb.org/View/2665583</link>
      <description><![CDATA[In this article, a novel high-frequency pulsating square-wave injection (HFPSI) strategy utilizing terminal voltage is proposed for sensorless control of dual three-phase permanent magnet synchronous motors (DTP-PMSMs) has been proposed. Different from the conventional high-frequency injection (HFI) methods with carrier current sensing, the proposed strategy utilizes the voltage induced by mutual inductance between two sets of windings as a carrier for positional information. Compared to traditional methods, the proposed approach effectively enhances the high-frequency (HF) components containing rotor position information, which facilitates the extraction and demodulation of HF voltage, thereby significantly improving sensorless control performance. At the same time, the acoustic noise can be attenuated by reducing the amplitude of the injection voltage. Besides, a switching strategy for this method has been presented in conjunction with a high-speed sensorless approach to ensure that the DTP-PMSM can achieve sensorless control across the entire speed range. This approach mitigates the influence of the sampling noise, thereby enhancing position accuracy. Finally, the validity of the proposed HFPSI strategy is experimentally confirmed.]]></description>
      <pubDate>Wed, 17 Jun 2026 16:13:03 GMT</pubDate>
      <guid>https://trid.trb.org/View/2665583</guid>
    </item>
    <item>
      <title>Enhancing Data Integrity for Autonomous Ships: A Multi-Level Detection Model Against Clean-Label Poisoning Attacks</title>
      <link>https://trid.trb.org/View/2679609</link>
      <description><![CDATA[With the deep integration of machine learning, the software-intensive architecture of autonomous ships, particularly their perception and decision-making modules, faces increasing challenges related to data integrity and resistance to tampering. Current safety frameworks are insufficient to address these risks, which are especially evident in ship object detection systems. Visual detection models, essential for safe navigation, remain highly vulnerable to clean-label poisoning attacks, in which training images are subtly manipulated without altering their labels. To address this gap, this study proposes a multi-level detection algorithm that integrates deep feature clustering with pixel-level anomaly inspection, providing progressive verification at the pixel, feature, and semantic levels. Experimental results on the MVDD13 maritime ship dataset demonstrate that even minimal poisoning can induce systematic and targeted detection errors without noticeably affecting overall model accuracy. For instance, at 5% poisoning ratio, overall precision and mAP@0.5 dropped to 0.89 and 0.86. The proposed algorithm effectively identifies poisoned data, thus enhancing the security and reliability of perception systems in autonomous ships. This approach offers a crucial step toward ensuring data integrity within the regulatory frameworks for safer autonomous navigation.]]></description>
      <pubDate>Wed, 17 Jun 2026 16:13:02 GMT</pubDate>
      <guid>https://trid.trb.org/View/2679609</guid>
    </item>
    <item>
      <title>Radar Detection of Birds Hazardous to Aircraft (Phase II)</title>
      <link>https://trid.trb.org/View/2703704</link>
      <description><![CDATA[A portable photographic radar recording system developed by the Technical Center Photographic Laboratory was evaluated to determine its capability to monitor and record bird movements as detected by an airport surveillance radar. This system was developed for use by U.S. Fish and Wildlife Service (USFWS) biologists for study of bird hazards problems at airports. The system was designed for photographing an automated radar terminal system (ARTS) II or ARTS III display with range and video selection available to the biologist. The design is such that interference with air traffic control equipment is kept to a minimum. Time and date documentation are provided. System operating instructions and technical data are contained in the appendices. Tests performed to demonstrate system capability included time lapse photography of various roosting area and migratory bird phenomena. These tests showed that the system was suitable for use by USFWS biologists for the study of bird movements hazards at airports.]]></description>
      <pubDate>Sun, 14 Jun 2026 16:17:13 GMT</pubDate>
      <guid>https://trid.trb.org/View/2703704</guid>
    </item>
    <item>
      <title>Automated rail spalling detection and quantification via pose rectification and enhanced deep semantic segmentation</title>
      <link>https://trid.trb.org/View/2701665</link>
      <description><![CDATA[Reliable detection and quantitative evaluation of rail surface spalling are challenged by strong reflections, worn textures, and irregular defect morphology. This paper presents an integrated framework covering image-based observation and three-dimensional damage assessment. A non-contact region extraction method identifies the geometrically preserved rail side using only grayscale statistics, providing a stable anchor for downstream processing. This region guides point cloud alignment and allows multi-segment scans to be rectified into a unified coordinate system with improved geometric consistency. On the rectified geometry, the proposed spalling segmentation network achieves an F-measure of 88.62% and an IoU of 85.01% on field-acquired data. The segmented regions are then mapped to the aligned point cloud to reconstruct their three-dimensional morphology and compute depth-, width-, and area-related indicators. The results demonstrate reliable defect geometry and stable pose alignment, offering an effective solution for automated rail condition assessment.]]></description>
      <pubDate>Fri, 12 Jun 2026 09:19:58 GMT</pubDate>
      <guid>https://trid.trb.org/View/2701665</guid>
    </item>
    <item>
      <title>Evolution-aware controllable diffusion for bridge coating damage image generation in automated detection</title>
      <link>https://trid.trb.org/View/2701660</link>
      <description><![CDATA[Automated visual inspection using UAVs and climbing robots is important for intelligent bridge maintenance. However, the scarcity of coating damage images, subtle features of early-stage defects, and difficulties in annotation in open scenarios severely restrict the performance of detection models. To this end, this paper proposes an evolution-aware controllable diffusion framework for generating high-fidelity bridge coating damage images. The method achieves accurate and controllable damage generation by decoupling appearance and spatial structure, utilizing saliency-guided attention, and applying a dual-branch residual denoiser. Meanwhile, based on optical-flow evolutionary priors, it simulates the degradation process of damages from mild to severe states. Experiments show the cross-category average LPIPS/SSIM/FID reaches 0.273/0.945/151.569. With data augmentation, the YOLO detector attains 80.2% ± 1.3%, 87.4% ± 0.9%, and 85.6% ± 0.8% AP for blistering, corrosion, and discoloration, which provides a practical route to support reliable coating condition assessment and to reduce the effort of collecting rare early-stage damages in bridge maintenance.]]></description>
      <pubDate>Fri, 12 Jun 2026 09:19:58 GMT</pubDate>
      <guid>https://trid.trb.org/View/2701660</guid>
    </item>
    <item>
      <title>Multimodal attention fusion for debonding detection in Ballastless tracks</title>
      <link>https://trid.trb.org/View/2706478</link>
      <description><![CDATA[To address the limited performance of single-modal automatic visual detection for debonding defects in CRTS I ballastless tracks under complex illumination conditions, this paper focuses on how to effectively exploit the complementary information between 2D and depth images to improve detection robustness and accuracy. A multimodal deep learning approach is developed based on the YOLOv8 framework, incorporating a multimodal attention fusion module with gated mutual guidance, adaptive fusion, and Coordinate Attention to enable bidirectional feature interaction and noise suppression. Experiments on a field-collected dataset demonstrate that: (1) the proposed model achieves an mAP@50 of 91.3%, representing a 36.2% improvement over single-modal baseline; (2) embedding the fusion module at intermediate and high-level feature layers effectively balances geometric detail with semantic representation; and (3) under shadow and low-light conditions, the proposed model significantly outperforms existing single-modal methods. This method breaks single-modal detection blind spots, supporting automated ballastless track inspection and maintenance automation.]]></description>
      <pubDate>Thu, 11 Jun 2026 09:29:17 GMT</pubDate>
      <guid>https://trid.trb.org/View/2706478</guid>
    </item>
    <item>
      <title>Second-level heterogeneous retired battery type identification using pulse-test-enabled federated learning with output-level privacy preservation</title>
      <link>https://trid.trb.org/View/2706538</link>
      <description><![CDATA[The imminent retirement of large numbers of lithium-ion batteries creates an urgent need for efficient reuse and recycling to realize lifecycle economic and environmental benefits. In this context, reliable battery type information is essential for downstream sorting, reuse evaluation, and recycling process selection, because different cathode chemistries are associated with different material values, processing requirements, and safety considerations. However, such information is often unavailable in practice due to long-term label degradation and restricted access to sensitive battery data, especially under non-independent and identically distributed (non-IID) conditions across stakeholders. To address this challenge, we propose a privacy-preserving federated learning framework for retired battery type identification using only second-level pulse-test data collected locally by clients. The framework incorporates an expert-weighted aggregation mechanism, in which client-specific autoencoders quantify local expertise through reconstruction error, while only classification probabilities and expert indices are transmitted for collaboration. This output-level design reduces privacy exposure by avoiding the exchange of raw data, and model parameters. Experiments on a heterogeneous retired-battery dataset containing 8 battery types, over 600 retired batteries, and 10,184 pulse-test records across LFP, LMO, NMC811, and NMC622 cells with nominal capacities ranging from 10 Ah to 68 Ah show that the proposed framework achieves an average classification accuracy of 96.3% across 100 stochastic non-IID partitioning scenarios. It consistently outperforms distributed aggregation baselines, including Average Aggregation, Class-Count Weighting, and Mahalanobis Distance Weighting, while remaining close to the centralized reference performance. By addressing label-distribution and data-quantity skew under a controlled non-IID setting, the proposed framework provides a methodological proof-of-concept for recovering chemistry-relevant battery type information from historically untraceable retired batteries, while external validation under independently collected datasets remains necessary for deployment-level assessment.]]></description>
      <pubDate>Thu, 11 Jun 2026 09:29:17 GMT</pubDate>
      <guid>https://trid.trb.org/View/2706538</guid>
    </item>
    <item>
      <title>A lightweight multi-scale gated expert network for ship-radiated noise recognition</title>
      <link>https://trid.trb.org/View/2710047</link>
      <description><![CDATA[Ship-radiated noise recognition remains challenging because segmented training recordings often contain redundant samples, single-resolution spectral inputs may not capture acoustic structures at different temporal scales, and static feature fusion cannot adjust scale contributions across samples. A lightweight multi-scale gated expert network is proposed for underwater ship-radiated noise recognition. The method integrates three task-oriented designs: a hard-random sample reconstruction strategy based on probe-estimated sample difficulty, a dual-resolution log-Mel input constructed with two hop sizes, and a gated multi-scale expert network with short-, medium-, and long-term temporal receptive-field branches. A Transformer encoder is used for temporal context modeling, and gating regularization is introduced to reduce near-uniform routing and single-branch dominance. Under a unified experimental protocol, the proposed model achieves accuracies of 98.43 ± 0.08% on DeepShip and 98.65 ± 0.07% on ShipsEar with 0.482M parameters. Additional MMSI-level vessel-disjoint evaluation on DeepShip, controlled acoustic perturbation tests, statistical significance analysis, and inference-cost evaluation are conducted to examine performance under stricter and degraded conditions. Ablation results indicate that sample reconstruction, dual-resolution representation, and gated multi-scale fusion each contribute to the final performance. The results support a favorable accuracy–complexity trade-off on the evaluated public datasets, while the conclusions remain limited to public benchmark protocols and controlled perturbation settings rather than open-sea deployment conditions.]]></description>
      <pubDate>Wed, 10 Jun 2026 16:38:53 GMT</pubDate>
      <guid>https://trid.trb.org/View/2710047</guid>
    </item>
    <item>
      <title>Classification of Bolt Corrosion Levels Combining Deep Learning and Multi-Feature Segmentation</title>
      <link>https://trid.trb.org/View/2712017</link>
      <description><![CDATA[Many bolts are installed in subway tunnels, making manual inspection prohibitively costly, and deep learning models face difficulties in segmenting extremely small corroded regions, which results in low detection efficiency. To address these challenges, this study proposes a corrosion grade classification algorithm for subway tunnel bolts based on deep learning and multi-feature segmentation, which directly outputs the corrosion grade of each bolt to enhance maintenance efficiency. First, the YOLOv8 framework is improved using multi-scale channel group shuffle convolution (MSCGSC) and focal loss (FL) to develop the YOLO-MF (MSCGSC + FL) model for preliminary detection of corroded bolts. Second, the VGG16 network is employed as the backbone of U-Net, and channel shuffle is applied after the encoder–decoder concatenation to eliminate background noise of bolts using the VGG + channel shuffle (VCS)-Net model. Finally, the fusion of segmentation features, spatial features, and clustering features enables the accurate segmentation and grading of tiny corroded areas. Experiment results demonstrate that YOLO-MF and VCS-Net achieve higher accuracy in corroded-bolt detection and background noise removal. Compared with other segmentation approaches, the multi-feature fusion segmentation method improves the intersection over union by 0.1623. The corrosion grade results are directly printed on the images, facilitating maintenance operations, reducing the workload of tunnel maintenance personnel, and improving tunnel maintenance efficiency.]]></description>
      <pubDate>Wed, 10 Jun 2026 09:06:00 GMT</pubDate>
      <guid>https://trid.trb.org/View/2712017</guid>
    </item>
    <item>
      <title>Hierarchical Cross-Attention Transformer for Remaining Voyage Time prediction using AIS and weather data</title>
      <link>https://trid.trb.org/View/2710170</link>
      <description><![CDATA[Accurate prediction of a vessel’s arrival time is essential for optimizing voyage schedules and port resource coordination within the Just-in-Time (JIT) arrival context. This study presents a Hierarchical Cross-Attention Transformer model for Remaining Voyage Time (RVT) prediction by integrating Automatic Identification System (AIS) trajectories, static vessel specifications, and maritime weather data. Unlike conventional Transformer-based approaches that process a voyage as a flat sequence, the proposed architecture partitions voyage sequences into temporal segments and employs a shared local Transformer encoder to capture intra-segment dynamics. A global Transformer encoder aggregates segment-level representations to model inter-segment dependencies, while a cross-attention mechanism dynamically fuses static vessel attributes with temporal features. The model was evaluated using a dataset of voyages bound for the Port of Busan and compared against several machine learning and deep learning baselines. Experimental results show that the proposed model achieved a Mean Absolute Error (MAE) of 29.35 min, outperforming the evaluated baseline models across various metrics. Furthermore, an arrival-time alignment analysis was conducted to examine the relationship between prediction accuracy and vessel residence time in low-speed regions. The results indicate that sequential RVT prediction has the potential to reduce unnecessary waiting periods, thereby supporting operational efficiency in maritime logistics.]]></description>
      <pubDate>Mon, 08 Jun 2026 15:48:01 GMT</pubDate>
      <guid>https://trid.trb.org/View/2710170</guid>
    </item>
    <item>
      <title>Study and Testing of Wheelset Wheel-Flat Identification Features through Axle-Box Vibration Measurements</title>
      <link>https://trid.trb.org/View/2676033</link>
      <description><![CDATA[Early detection of wheelset defects is essential for ensuring railway safety. Wheelset condition monitoring can provide continuous information about the health of the system, thus avoiding time-consuming and expensive operations such as periodic inspections. This work deals with the study of railway wheelset wheel-flat identification features based on vibration signals from axle-box measurements. The aim is to obtain a simple and straightforward solution that can be easily implemented on a complete autonomous on-board sensor for wheelset defect prediction. Numerical simulations, by coupling a multi-body model of a coach with a new approximation of a wheel-flat model, were run in order to estimate the nature of the problem and the technical acquisition characteristics needed for a sensor node to be installed on real trains. Then, experimental campaigns were carried out on a wheelset test bench with defects artificially created to validate the presented methodology based on time domain feature extraction. A signal processing technique, which does not require the aid of any other hardware to obtain the revolution speed, is proposed. The methodology allows clear detection of wheel flats starting from 30 mm, especially at lower speeds. Even when considering the influence of wear, high defect conditions remain easily distinguishable.]]></description>
      <pubDate>Mon, 08 Jun 2026 08:38:12 GMT</pubDate>
      <guid>https://trid.trb.org/View/2676033</guid>
    </item>
    <item>
      <title>An Electrified Railway Catenary Component Anomaly Detection Frame Based on Invariant Normal Region Prototype With Segment Anything Model</title>
      <link>https://trid.trb.org/View/2665559</link>
      <description><![CDATA[The existing drone-based detection systems for electrified railway catenary support components face the following challenges: 1) the background of aerial images of the catenary support component is complex; 2) a unified model struggles to handle anomaly detection for various component types; and 3) there is inconsistency in the normal region feature information between training and testing images. To address these issues, this article proposed a novel adaptive anomaly detection framework for catenary support component an anomaly detection model based on invariant normal region prototype extraction (INRP-Ader). First, we proposed a new segmentation model (CSC-SAM) that embeds key catenary component location information to extract foreground images. Next, we design an anomaly detection model that directly extracts invariant normal region prototype (INRP) features from the test images. This model includes an INRP extractor constrained by INRP smoothness loss and an INRP-guided decoder, aimed at solving the problem of inconsistency in normal region features between training and testing images and the challenge of adapting a unified model to multiple component types. In addition, a soft mining loss is introduced to further optimize the training process of the multiclass anomaly detection model. Finally, we established a real-world catenary support component dataset, catenary system component UAV dataset (CSCUD), collected by drones, and achieved detection performance of 99.1/98.2/98.2 in image-level metrics (I-AUROC/I-AP/I-F1) and 96.4/37.1/55.7 in pixel-level metrics (P-AUROC/P-AP/P-F1). The proposed method outperforms traditional methods by approximately 0.4%–16.6% across various metrics.]]></description>
      <pubDate>Mon, 08 Jun 2026 08:38:12 GMT</pubDate>
      <guid>https://trid.trb.org/View/2665559</guid>
    </item>
    <item>
      <title>Application of Image Analysis ‘Engine’ to Night Vision Images of Railway Track Periphery</title>
      <link>https://trid.trb.org/View/2675871</link>
      <description><![CDATA[We developed a method for capturing and analyzing images captured at night and verified the feasibility for limit measurement and difference detection of the environment. The result confirmed that by combining infrared light projectors, it is possible to capture a clear image of the rail periphery. We also developed a method for measuring the distances around the platform and examined the measurement accuracy. The average error was less than 20 mm. Furthermore, we confirmed that the difference detection can be properly performed even for night vision images.]]></description>
      <pubDate>Fri, 05 Jun 2026 16:41:22 GMT</pubDate>
      <guid>https://trid.trb.org/View/2675871</guid>
    </item>
    <item>
      <title>Derailment Detection for Wagons Using Mechanical Contact Sensors</title>
      <link>https://trid.trb.org/View/2675868</link>
      <description><![CDATA[In long freight trains, if a wagon far from the locomotive derails, detection of the derailment may be delayed, potentially allowing the train to continue running with the derailed wagon. We investigated a derailment detection method using mechanical contact sensors to enable early detection of wagon derailments. In this study, the installation position of the contact sensors was selected based on results from past derailment investigations and simulation results. Derailment detection tests were then conducted derailing a bogie equipped with contact sensors on an actual track. These tests confirmed that the contact sensors can detect derailment immediately after it occurs.]]></description>
      <pubDate>Fri, 05 Jun 2026 16:41:22 GMT</pubDate>
      <guid>https://trid.trb.org/View/2675868</guid>
    </item>
  </channel>
</rss>