<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/common/TRIS Suite/feeds/RSS.aspx?s=PHNlYXJjaD48cGFyYW1zPjxwYXJhbSBuYW1lPSJzcGVjaWZpY3Rlcm1zIiB2YWx1ZT0iOTI2wqE1MTA1NcKhNTEwODTCoTI5NznCoTExMDUwN8KhOTQ0NzXCoTUwNzDCoTU0MjQiIC8%2BPHBhcmFtIG5hbWU9InN1YmplY3Rsb2dpYyIgdmFsdWU9Im9yIiAvPjxwYXJhbSBuYW1lPSJ0ZXJtc2xvZ2ljIiB2YWx1ZT0ib3IiIC8%2BPHBhcmFtIG5hbWU9ImxvY2F0aW9uIiB2YWx1ZT0iMCIgLz48L3BhcmFtcz48ZmlsdGVycyAvPjxyYW5nZXMgLz48c29ydHM%2BPHNvcnQgZmllbGQ9InB1Ymxpc2hlZCIgb3JkZXI9ImRlc2MiIC8%2BPC9zb3J0cz48cGVyc2lzdHM%2BPHBlcnNpc3QgbmFtZT0icmFuZ2V0eXBlIiB2YWx1ZT0icHVibGlzaGVkZGF0ZSIgLz48L3BlcnNpc3RzPjwvc2VhcmNoPg%3D%3D" 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>Metaheuristic and reinforcement learning techniques for solving the vehicle routing problem: A literature review</title>
      <link>https://trid.trb.org/View/2785034</link>
      <description><![CDATA[The vehicle routing problem (VRP) remains a pivotal challenge in combinatorial optimization, where the objective is to determine optimal routes for a fleet of vehicles serving geographically distributed customers under specific constraints such as capacity, time windows, and demand variability. Over the decades, a diverse spectrum of solution methodologies has emerged to address the NP-hard nature of this problem and its numerous real-world variants. This review critically examines the intersection of metaheuristics (MHs) and reinforcement learning (RL), with a particular focus on hybrid frameworks that exploit the complementary strengths of both paradigms: the broad exploratory power of MHs and the adaptive, experience-driven decision-making of RL. While existing literature has predominantly explored the role of metaheuristics in fine-tuning RL parameters and architectures, the inverse, using RL to enhance metaheuristic search adaptivity and real-time decision-making, remains markedly underexplored. This paper presents a bidirectional analysis of their integration, examining (1) how metaheuristics enhance RL-based solutions and (2) how reinforcement learning improves the efficiency and adaptability of metaheuristics. Drawing on a systematic review of 279 relevant studies across seven major academic databases, we find that only 13.2% (37 papers) explicitly integrate both techniques into unified hybrid frameworks, rather than merely benchmarking one against the other. Based on this curated set of studies, we propose a novel classification framework that organizes hybrid MH-RL approaches into three distinct categories: MH-based RL/improving MHs, enhancing RL, and dynamic optimization, analyzing their underlying principles, mathematical formulations, methodologies, and contributions. Furthermore, we discuss key limitations of current methods, including high computational costs and standardization challenges, and outline future research avenues, offering insights into emerging trends and the potential of this hybridization to advance VRP optimization.]]></description>
      <pubDate>Mon, 05 Oct 2026 09:07:30 GMT</pubDate>
      <guid>https://trid.trb.org/View/2785034</guid>
    </item>
    <item>
      <title>High-Speed rail expansion and avian Biodiversity Decline: Evidence from a Hub-Link Perspective</title>
      <link>https://trid.trb.org/View/2785056</link>
      <description><![CDATA[Although the economic effects of high-speed rail (HSR) are well documented, its implications for avian biodiversity remain unclear. Using a county-level panel for China from 2002 to 2020, we combine effort-adjusted indices from citizen science bird records with geospatial data on HSR lines and stations and estimate treatment effects using a Double Machine Learning (DML) framework. A hub-link decomposition indicates that the negative estimate for avian biodiversity is concentrated in counties with HSR stations, whereas pass-through corridors have no detectable average effect at the county scale. The mechanism analysis is consistent with increased human activity and land conversion around stations. A micro-spatial buffer analysis further shows that the avian estimates are most negative within 1 km of stations and diminish with distance. These findings suggest that ecological assessments of HSR projects should consider not only rail alignments but also station-area land development.]]></description>
      <pubDate>Mon, 05 Oct 2026 09:07:30 GMT</pubDate>
      <guid>https://trid.trb.org/View/2785056</guid>
    </item>
    <item>
      <title>Housing type mix as a lever for sustainable urban mobility</title>
      <link>https://trid.trb.org/View/2785053</link>
      <description><![CDATA[Land use-transport interactions are well documented in literature using urban form metrics including density, land–use mix, and destination accessibility, but rarely housing type shares. This study demonstrates the overlooked explanatory and causal relevance of housing typology, specifically the single–family housing share, for four mobility outcomes: car travel, transit travel, car ownership, and CO₂ emissions. Using postcode–level data for 13 cities in Germany and France, we compare the role of urban form metrics across linear regression, gradient–boosted decision trees, artificial neural networks, causal discovery, and double machine learning methods. Single–family share consistently outperforms conventional urban form indicators in explaining each outcome. Given its strong predictive power, causal relevance, and ease of measurement, housing typology offers substantial potential for modelling land use-transport interactions and provides a promising cross–sectoral focus for climate and mobility policy.]]></description>
      <pubDate>Mon, 05 Oct 2026 09:07:30 GMT</pubDate>
      <guid>https://trid.trb.org/View/2785053</guid>
    </item>
    <item>
      <title>Computer Vision-Based Contactless Estimation of Asphalt Pavement Surface Friction</title>
      <link>https://trid.trb.org/View/2784622</link>
      <description><![CDATA[This study developed a contactless and automated technique to estimate hot-mix asphalt pavement surface friction by combining computer vision image–based 3D reconstruction with machine learning. Existing continuous friction estimation equipment, including high-speed laser scanners, are costly to purchase and operate—limiting routine, frequent monitoring. This study addressed this limitation through the development and utilization of cost-effective and scalable equipment. A Structure-from-Motion computer vision framework was developed to reconstruct pavement macrotexture in three dimensions and compute mean profile depth (MPD). Ground-control-based measurements agreed with a reference laser scanner within 15% across all test locations. A high-speed prototype using industrial cameras and a real-time kinematic positioning system was built. It was able to collect data up to 35 mi/hr and perform 3D reconstruction on cloud environments, increasing processing speed. The prototype was validated against a laser scanner at a controlled test track. The results demonstrated that contactless texture measurement and friction prediction are feasible and provide a foundation for more frequent, lower-cost monitoring of pavement surface characteristics, supporting proactive pavement management, safety screening, and more cost-effective maintenance decisions. The developed device measures texture and estimates friction for $61/mile, about 65% less than the $140/mile needed to collect both with current methods, and it costs only $73,500 to buy, compared with about $1,100,000 for a combined laser scanner and a transverse profiler. In addition, a data-driven model was developed to predict sideways friction measured by a Sideway-force Coefficient Routine Investigation Machine using pavement texture, traffic, roadway geometry, operating speed, and surface temperature. Three datasets collected in Florida and Illinois were used to evaluate the influence of dataset size, available features, and data distribution. Among the tree-based ensemble models evaluated, the Extra Trees model performed best, explaining approximately 85% of the variance, with MPD, traffic, operating speed, and surface temperature identified as the most influential variables. Finally, a finite element framework coupling a 3D tire model, texture-based friction model, and water film was developed to evaluate hydroplaning. The framework demonstrated that coarser macrotextures resulted in relatively better tire-pavement contact at higher speeds because of their improved surface drainage capacity.]]></description>
      <pubDate>Mon, 05 Oct 2026 08:56:38 GMT</pubDate>
      <guid>https://trid.trb.org/View/2784622</guid>
    </item>
    <item>
      <title>Roadmap for Virtual Traffic Light Control at Intersections Using Connectivity, AI and Digital Twins</title>
      <link>https://trid.trb.org/View/2786712</link>
      <description><![CDATA[Vehicle automation, connectivity, artificial intelligence (AI), and advanced sensing enable new approaches to intersection management. This project investigates a virtual traffic light (VTL) framework that over time can replace conventional signalized intersections, which carry high capital, maintenance, and operating costs. In the proposed system, connected and automated vehicles (CAVs), along with connected pedestrians and bicyclists, communicate continuously with a virtual control agent that maintains real-time estimates of their positions, speeds, and intended trajectories. Using these data, the agent computes individualized right-of-way assignments—equivalent to virtual green, yellow, and red indications—and transmits them directly via vehicle-to-everything (V2X) communication to each user. By coordinating movements through communication rather than fixed infrastructure, the system can improve traffic efficiency, reduce delays, enhance safety, and lower costs. An additional advantage of VTL systems is their flexibility: control logic and policies can be updated through software, enabling rapid, low-cost adaptation compared with conventional traffic signals. Because full connectivity will be achieved gradually, the research first establishes upper-bound performance under ideal, fully connected conditions without failures, providing quantitative benchmarks. It then develops a roadmap for transitioning from today’s traffic control to increasingly connected environments, accommodating mixed traffic with varying levels of connectivity. The results will guide phased deployment strategies aligned with USDOT priorities in safety, mobility, sustainability, and cost-effective infrastructure modernization.

]]></description>
      <pubDate>Mon, 05 Oct 2026 13:32:38 GMT</pubDate>
      <guid>https://trid.trb.org/View/2786712</guid>
    </item>
    <item>
      <title>Machine Learning-based Analysis for Railway Track Maintenance</title>
      <link>https://trid.trb.org/View/2784735</link>
      <description><![CDATA[Understanding and mitigating track degradation rates (TDRs) is crucial for ensuring the safety and efficiency of railway systems, as well as for controlling maintenance costs and preserving ride quality. The Federal Railroad Administration (FRA) sponsored the University of Massachusetts Amherst and Loram Technologies, Inc. to extensively analyze TDR over 280 miles of passenger revenue track using data from 2011 to 2021. Using a large dataset was pivotal, as it enhanced the robustness of the analysis, allowing for more accurate and reliable conclusions. The researchers used ground penetrating radar (GPR) and light detection and ranging (LiDAR) along with geometry data to examine subsurface and drainage conditions of the tracks. The project employed two distinct methods to estimate TDR, each offering unique insights and applicability based on available historical data and desired precision. These complementary approaches provide the industry with versatile tools for estimating TDR and improving maintenance scheduling, emphasizing their reliability and adaptability.]]></description>
      <pubDate>Sat, 03 Oct 2026 18:00:47 GMT</pubDate>
      <guid>https://trid.trb.org/View/2784735</guid>
    </item>
    <item>
      <title>Virtual Inspection Systems for Highway Construction Projects: A Guide</title>
      <link>https://trid.trb.org/View/2786112</link>
      <description><![CDATA[This report presents a guide for assessing the readiness of digital technologies to support remote virtual inspection (RVI) activities for highway construction projects. This guide was developed around an analytical methodology that incorporates stakeholder interviews, case studies, and a peer review workshop. This guide will be of immediate interest to highway construction engineers, project managers, inspectors, and surveyors seeking to integrate these technologies effectively to support inspection requirements.]]></description>
      <pubDate>Sat, 03 Oct 2026 18:00:47 GMT</pubDate>
      <guid>https://trid.trb.org/View/2786112</guid>
    </item>
    <item>
      <title>Virtual Inspection Systems for Highway Construction Projects: Developing a Guide</title>
      <link>https://trid.trb.org/View/2786113</link>
      <description><![CDATA[This report, developed under NCHRP Project 10-112, offers practical guidelines for implementing digital technologies to support remote virtual inspection (RVI) in highway infrastructure projects. It evaluates the effectiveness, opportunities, and challenges associated with a range of RVI tools, including mobile devices, artificial intelligence (AI), machine learning (ML), lidar, small uncrewed aerial systems (sUAS), remote imaging, augmented reality and virtual reality (AR/VR), and closed-circuit television (CCTV). Drawing on an extensive literature review, a nationwide survey of state departments of transportation (DOTs), and a series of case studies, the report presents a structured methodology for assessing technologies and integrating them into existing inspection workflows. The findings highlight the potential of RVI technologies to improve inspection accuracy, efficiency, and safety. Survey responses indicate that mobile devices are currently the most widely used tool, while more advanced technologies like lidar and sUAS are emerging but less commonly applied. The report also identifies key barriers—such as cost, training, and infrastructure needs—and outlines future directions for broader adoption. A flexible technology assessment framework is introduced to help agencies evaluate the readiness and suitability of RVI technologies based on factors such as technical maturity, operational fit, cost-effectiveness, and implementation challenges. The framework can be tailored to meet the specific needs of individual agencies and project goals. Five case studies from DOT-led projects were used to validate the proposed approach, demonstrating general alignment with agency decision-making and revealing insights for further refinement. A collaborative workshop with DOT and industry stakeholders also informed the development of the Guide, emphasizing the importance of technology champions, return on investment (ROI) tracking, and data management planning. The report concludes with a set of actionable recommendations covering standardization of RVI procedures, workforce development, strategic planning, and addressing implementation hurdles. It also explores future trends, including the integration of AI, ML, and Internet of Things (IoT) sensors to enable more automated, real-time inspection workflows. Overall, the report underscores the transformative potential of RVI technologies in modernizing highway construction inspection. It provides a practical foundation to help transportation agencies navigate the shift toward more digital, efficient, and data-driven inspection practices.]]></description>
      <pubDate>Sat, 03 Oct 2026 18:00:47 GMT</pubDate>
      <guid>https://trid.trb.org/View/2786113</guid>
    </item>
    <item>
      <title>Adaptive Region Distillation: A lightweight knowledge distillation method for ship fuel consumption prediction</title>
      <link>https://trid.trb.org/View/2782414</link>
      <description><![CDATA[Accurate ship fuel consumption prediction is essential for energy efficiency optimization, route planning, and emission reduction. However, interactions among operational and environmental factors produce local fluctuations and heteroscedasticity across operating conditions, making it difficult for lightweight onboard models to preserve predictive structures learned by high-capacity models.Existing regression knowledge distillation methods mainly align point predictions or global predictive distributions, providing limited representation of local and relative structure in continuous output spaces. To address this limitation, this paper proposes Adaptive Region Distillation (ARD), a knowledge distillation framework with region decoupling for ship fuel consumption regression. ARD constructs adaptive proxy distributions around observed targets and decouples knowledge transfer into target-neighborhood and complementary-neighborhood objectives. ARD was evaluated on operational records from two container ships and one cargo ship using MLP and LSTM backbones. Relative to the standalone Student, ARD increased mean R² by 0.0308–0.0870 for MLP and 0.0070–0.0954 for LSTM. Relative MAPE reductions were 16.2%–33.5% for MLP, while absolute reductions for LSTM were 0.7256–2.1721 percentage points. Ablation and sensitivity analyses characterized component contributions and hyperparameter responses. ARD introduces no additional parameters or branches at inference time, retaining parameter compression ratios of 167.67× and 252.17× for the MLP and LSTM students, respectively.]]></description>
      <pubDate>Fri, 02 Oct 2026 16:54:22 GMT</pubDate>
      <guid>https://trid.trb.org/View/2782414</guid>
    </item>
    <item>
      <title>A hierarchical spatio-temporal reinforcement learning framework for cooperative multi-USV search and interception</title>
      <link>https://trid.trb.org/View/2782404</link>
      <description><![CDATA[Multi-USV cooperative search and target interception holds value in maritime security, yet existing methods face three challenges: multi-timescale decision coupling, insufficient heterogeneous observation fusion, and inefficient exploration under sparse rewards. This paper proposes a multi-USV decision-making framework based on Hierarchical Spatio-Temporal Proximal Policy Optimization (HST-PPO) with three contributions. A three-layer spatio-temporal decision architecture spanning strategic, tactical, and execution layers is coupled via soft goal passing and trained end-to-end. A Hierarchical Causal Observation Fusion Network (HCOF-Net) uses causal mask attention and temporal decay encoding to adaptively fuse four categories of observations. A Distributed Random Network Distillation exploration mechanism (DRND-NovelD) combines ensemble novelty estimation with a multi-agent diversity reward to accelerate convergence under sparse rewards. HST-PPO is validated across Python simulation, Unity high-fidelity simulation, and real-world lake trials, and benchmarked against ten baselines spanning six non-hierarchical MARL methods and four hierarchical RL methods adapted to the same setting. HST-PPO achieves a Search Coverage Rate of 82.3% and Interception Success Rate of 80.0%, surpassing the strongest baseline by 1.7 and 1.1 percentage points and outperforming every hierarchical baseline. In real-world trials, HST-PPO retains 96.2% and 81.3% of its simulated Search Coverage Rate and Interception Success Rate, validating its zero-shot sim-to-real transfer capability.]]></description>
      <pubDate>Fri, 02 Oct 2026 16:54:22 GMT</pubDate>
      <guid>https://trid.trb.org/View/2782404</guid>
    </item>
    <item>
      <title>A radar–vision fusion approach for ship monitoring in bridge waterways using weakly supervised cross-modal representation learning</title>
      <link>https://trid.trb.org/View/2782369</link>
      <description><![CDATA[Frequent ship–bridge collisions pose serious risks to infrastructure, public safety, and economic assets. Reliable ship perception is critical for timely risk identification in bridge waterways. Radar–vision fusion enables complementary ship sensing capabilities beyond single-sensor monitoring using cameras or radars alone. However, existing rule-based fusion methods exhibit limited applicability in continuous waterway monitoring due to poor generalisation and labour-intensive calibration. This study proposes a weakly supervised representation learning framework that achieves embedding-level alignment with reduced manual supervision to improve the robustness of radar–vision association-based ship monitoring. A time window-based pseudo-labelling strategy is introduced to guide a Transformer-based dual encoder for learning shared radar–vision representations, with a novel ordinal loss incorporated into contrastive learning to enforce physical consistency. The framework is validated on a real-world bridge waterway using a field-deployed radar–camera system, achieving an F1-score of 95.7% and outperforming state-of-the-art learning-based fusion methods in the traffic sensing domain. The proposed approach can provide reliable and automated situational awareness of bridge waterway traffic to support decision-making in navigational safety management systems.]]></description>
      <pubDate>Fri, 02 Oct 2026 16:54:22 GMT</pubDate>
      <guid>https://trid.trb.org/View/2782369</guid>
    </item>
    <item>
      <title>Artificial Intelligence (AI) for Integrated Safety Assessment Model (ISAM)</title>
      <link>https://trid.trb.org/View/2780457</link>
      <description><![CDATA[This project aims to advance risk-informed decision-making at the Federal Aviation Administration (FAA) by developing machine learning techniques for extracting safety-relevant information from incidents and accident reports and enhancing large language model (LLM) capabilities for navigating aviation regulatory text. The work addresses two primary challenges: the labor-intensive process of manually extracting safety data from Aviation Safety Reporting System (ASRS) and National Transportation Safety Board (NTSB) narratives for use in the Integrated Safety Assessment Model (ISAM), and the inability of general-purpose language models to accurately interpret the hierarchical structure and cross-references within Title 14 of the Code of Federal Regulations (CFR). To improve safety data extraction, the project developed two datasets. The first is a question answering (QA) dataset designed to enable precise supervision for fine-tuning LLMs to extract events, contributing factors, and operational conditions from narrative text. The second consists of synthetic safety reports generated using a fine-tuned LLM informed by a domain-specific knowledge graph (KG) built from FAA aircraft data to eliminate aircraft specification errors and improve multilabel classification performance. To enhance regulatory reasoning, the project built a regulation-aware knowledge graph and developed a retrieval- augmented generation (RAG) system with improved retrieval precision for semi-structured regulatory content. An evaluation framework was also implemented by adapting retrieval-augmented generation assessment (RAGAS) metrics for regulatory contexts and introducing a custom answer completeness metric to address the “missing zone problem” of incomplete retrieval. Finally, the FAA-provided Daedalus multi-agent RAG system was migrated to local infrastructure using the Model Context Protocol (MCP) and extended with a completeness-evaluation agent that provides actionable diagnostics by distinguishing between retrieval and generation incompleteness.]]></description>
      <pubDate>Fri, 02 Oct 2026 16:53:02 GMT</pubDate>
      <guid>https://trid.trb.org/View/2780457</guid>
    </item>
    <item>
      <title>Analyzing Pre- and Post-Coastal Hazard Pavement Conditions to Optimize Response Strategies for Coastal Infrastructure Resilience</title>
      <link>https://trid.trb.org/View/2780496</link>
      <description><![CDATA[Coastal regions are increasingly exposed to hazards such as storm surge, flooding, erosion, and sea-level rise, which pose significant risks to civil infrastructure, particularly transportation assets. This study evaluated the impacts of Hurricane Harvey on highway pavement performance in the Houston area using historical Pavement Management Information System (PMIS) data, statistical analyses, and machine learning methods. Pre- and post-Harvey pavement conditions were compared across Asphalt Concrete Pavement (ACP), Jointed Concrete Pavement (JCP), and Continuously Reinforced Concrete Pavement (CRCP), with consideration of pavement distress types, severity levels, and condition changes. A Wilcoxon signed-rank test was first applied to evaluate differences between pre- and post-Harvey pavement conditions and identified a statistically significant network-level decline following the hurricane. Multiple Linear Regression (MLR) and Random Forest (RF) models were subsequently developed to evaluate factors associated with post-Harvey pavement performance and condition loss. The modeling results identified pre-Harvey pavement condition, pavement type, traffic loading, and environmental factors as important predictors of post-Harvey performance, although their relative importance varied depending on the response variable evaluated. The RF models provided improved predictive performance and captured nonlinear relationships among the contributing factors, while the MLR models provided complementary information regarding the direction and magnitude of statistical associations. The results demonstrated differences in hurricane-related performance among pavement types, with CRCP generally exhibiting greater resilience, ACP experiencing varying levels of localized and distributed deterioration, and JCP exhibiting greater vulnerability to hurricane-related impacts. Overall, the findings demonstrate that pavement response to extreme weather events is influenced by a combination of pre-existing pavement condition, structural characteristics, traffic loading, and environmental exposure, providing a basis for improved pavement resilience assessment and risk-informed infrastructure management in hurricane-prone regions.]]></description>
      <pubDate>Fri, 02 Oct 2026 16:53:02 GMT</pubDate>
      <guid>https://trid.trb.org/View/2780496</guid>
    </item>
    <item>
      <title>Digitalisation and public transport</title>
      <link>https://trid.trb.org/View/2786106</link>
      <description><![CDATA[Digitalisation plays an important role for society and public transport as it can influence all dimensions of sustainability – social, ecological and economic. For instance, digitalisation can influence public transport resource efficiency, thereby making it possible to provide better service and save on resources, which includes reducing emissions. Furthermore, digitalisation can influence groups and individuals in different ways. While digitalisation can support passengers, it can also exclude people who are less used to using digital services such as buying tickets with a smartphone, or who are unable to use such services. As digitalisation is typically based on access to data, there is also a risk of violations of personal privacy due to information about individuals being created and stored without their consent. Digitalisation can also affect which stakeholders have access to data, and to what extent, which allows some stakeholders to gain competitive advantages over others.]]></description>
      <pubDate>Fri, 02 Oct 2026 09:37:39 GMT</pubDate>
      <guid>https://trid.trb.org/View/2786106</guid>
    </item>
    <item>
      <title>Experimental study on ship attitude prediction under wave conditions based on ResCNN-BiLSTM-attention hybrid algorithm</title>
      <link>https://trid.trb.org/View/2782453</link>
      <description><![CDATA[To address the strong nonlinearity and time-varying characteristics of ship attitude responses under wave excitation and the demand for accurate long-horizon prediction, a wave-tank experiment was conducted using a self-developed catamaran under a moderate wave condition with an average wave height of 1.02 m. Roll and pitch responses, together with the corresponding angular velocities and ship-body linear accelerations, were measured at the bow and stern. After detrending, band-pass filtering, downsampling, and sliding-window sample construction, a ResCNN-BiLSTM-Attention model was developed to predict roll and pitch responses over a future horizon of 100 s using a 60-s historical window. The proposed model integrates a residual one-dimensional convolutional encoder for local dynamic-feature extraction, a BiLSTM network for temporal dependency modeling, and a temporal attention mechanism for adaptive aggregation of dynamically important historical information. Its performance was compared with those of a conventional LSTM and a representative Attn-BiLSTM-GCNN hybrid model. Ablation experiments and ten independent training runs were further conducted to evaluate the contributions of individual architectural components and the robustness of the performance improvements. The proposed model achieved the lowest MSE, RMSE, and MAE in all four roll and pitch prediction tasks. Compared with the conventional LSTM, the average RMSE, MAE, and MSE were reduced by approximately 36.93%, 42.76%, and 59.43%, respectively. Relative to Attn-BiLSTM-GCNN, the average MSE, RMSE, and MAE were further reduced by approximately 16.76%, 12.15%, and 14.69%, respectively. The ablation results showed that incorporating both residual feature preservation and temporal attention reduced the average RMSE from 0.2738 for CNN-BiLSTM to 0.2353 for the proposed model. Moreover, the proposed model achieved the lowest mean RMSE across all four prediction tasks over ten independent runs. These results demonstrate the effectiveness and robustness of the proposed method for long-horizon multi-step catamaran attitude prediction under the investigated wave condition, while further validation under different sea states and vessel conditions remains necessary.]]></description>
      <pubDate>Thu, 01 Oct 2026 11:44:53 GMT</pubDate>
      <guid>https://trid.trb.org/View/2782453</guid>
    </item>
  </channel>
</rss>