<?xml version="1.0" encoding="utf-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>TRB Publications Index</title><link>http://pubsindex.trb.org/</link><atom:link href="http://pubsindex.trb.org/common/TRIS Suite/feeds/rss.aspx" rel="self" type="application/rss+xml" /><description></description><language>en-us</language><copyright>Copyright © 2015. 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>TRB Publications Index</title><url>http://pubsindex.trb.org/Images/PageHeader-wTitle.png</url><link>http://pubsindex.trb.org/</link></image><item><title>Optimized Laboratory Fabrication of Small-Specimen Geometry for Streamlining Dynamic Modulus and Cyclic Fatigue Testing of Asphalt Mixtures</title><link>http://pubsindex.trb.org/view/2709230</link><description><![CDATA[The asphalt community is focused on the paradigm shift in mixture design from the volumetrics to an optimization procedure based on performance testing called balanced mixture design. Streamlining performance testing to obtain index properties quickly and using a smaller quantity of materials is critical for the successful implementation. This paper aims to streamline dynamic modulus (|E*|) and cyclic fatigue testing by optimizing the number of 38 mm diameter specimens extracted from a single 150 mm diameter Superpave gyratory-compacted (SGC) specimen. The current provisional standard methods require vertical coring of four small specimens from a single SGC specimen. In this study, two sets of testing specimens were fabricated by coring four and five small specimens from each SGC specimen. The success rate in meeting target air voids, the |E*| analysis, and the cyclic fatigue results including cyclic fatigue index parameter (Sₐₚₚ) values were compared between the two sets of specimens. No significant or consistent differences were observed in performance testing results. Furthermore, innovative image analysis and microscopy techniques were used to study air voids distribution and aggregate structure within each specimen and to further validate the proposed coring pattern. Based on these findings, coring five 38 mm diameter testing specimens from one SGC sample is suggested to run |E*| and cyclic fatigue tests. This proposed modification to AASHTO TP 132 and TP 133 may save technicians’ time and allows for the optimal use of materials. The latter may become a significant saving when integrating these methods with laboratory long-term aging protocols and forensic studies.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709230</guid></item><item><title>Mechanistic Analysis and Design Framework for Geosynthetic Stabilized Unpaved Roads</title><link>http://pubsindex.trb.org/view/2709229</link><description><![CDATA[Geosynthetics provide mechanical stabilization benefits to paved or unpaved roads through lateral restraint of unbound aggregate particles and bearing capacity improvement over weak subgrades. The current state of the art incorporating geosynthetics into paved or unpaved road design involves conducting proper elastic layered system mechanistic analysis to determine the improvement of aggregate layer stiffness for increased traffic capacity or reduction in aggregate layer thickness. This paper presents a mechanistic analysis and design pipeline for determining the required aggregate thickness via the finite element (FE) modeling approach. An advanced FE analysis tool, C-FLEX, was employed to analyze axisymmetric multilayered unpaved road structures, accounting for the nonlinear stress-dependent behavior of unbound aggregates. The modulus enhancements were quantified for 10 different geosynthetics using the latest Bender Element sensor technology in both triaxial and large-scale tests conducted on typical dense-graded base aggregates. They were then incorporated into base course stiffness characterization via a sublayering approach for the unpaved road comprising aggregate base placed over soft subgrade. Both the measured enhanced moduli and the the extent of geosynthetic influence zones were adequately established in the sublayering approach. Further, sensitivity analysis was conducted for different aggregate modulus models and different sublayer structures, which verified the proposed design pipeline to provide satisfactory results. The method was also compared with the Giroud and Han method, which revealed the inherent difference in the two methods, given that the design here is based on the critical pavement responses and subgrade strength, while the Giroud and Han method also incorporated the field data with performance evaluation.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709229</guid></item><item><title>Enhancing Data Accessibility through Automated Personally Identifiable Information De-Identification in Crash Narratives</title><link>http://pubsindex.trb.org/view/2709228</link><description><![CDATA[Unstructured crash narratives in police reports contain rich textual information that can uncover key insights into crash circumstances, such as contributing factors and driver behavior, that are often missing from the structured fields of crash data. However, the presence of personally identifiable information (PII) within these narratives, and the lack of scalable, domain-specific redaction tools, limit their broader use because of privacy concerns and legal restrictions. To address this challenge, a scalable, privacy-preserving pipeline for automated PII de-identification from crash narratives was developed and evaluated. The proposed method utilizes a generalist model for named entity recognition using bidirectional transformer (GLiNER), which is known for its strong zero-shot, few-shot, and fine-tuned performance across diverse entity types. The model was fine-tuned on a manually annotated training set to adapt it to the crash narrative domain. It was found that combining this fine-tuned named entity recognition model with a rule-based post-processing module improved PII detection performance by resolving span misalignments and recovering entities that were initially missed. Evaluation on a test set achieved an F1 score above 80%, particularly for frequent PII categories such as names and addresses. Post-processing further reduced false negatives by 32%. The pipeline was developed and tested on local machines to ensure data confidentiality. Additionally, the workflow supports accessibility and future use through GLiNER-Studio, a user-friendly tool that enables non-programmers to fine-tune models on new datasets. This study contributes a practical solution to the need for automated PII de-identification in transportation safety data, enabling secure data sharing and ethical analytics for research and policymaking.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709228</guid></item><item><title>Pickup and Delivery Problem with Synchronous and Asynchronous Transshipment for Q-Commerce Delivery</title><link>http://pubsindex.trb.org/view/2709227</link><description><![CDATA[This study explores ways to improve delivery efficiency and reduce costs in the rapidly growing quick commerce (Q-commerce) market, which is expanding because of advancements in Internet of Things technology and the rise of contactless consumption. However, it faces significant challenges, such as high transportation costs and inefficient vehicle utilization. To address these challenges, a novel delivery system is proposed that simultaneously incorporates synchronous and asynchronous transshipment. The proposed system, the Pickup and Delivery Problem with two types of transshipment, is formulated as a mathematical optimization model. Since the problem is Nondeterministic Polynomial-time hard (NP-hard), implying that finding an optimal solution is computationally intensive as the problem scale increases, a two-phase heuristic algorithm is developed that combines adaptive large neighborhood search metaheuristic with transshipment scheduling. In Phase 1, adaptive large neighborhood search is employed to improve the initial Pickup and Delivery Problem solution. In Phase 2, the two transshipments are integrated into the improved solution. Experimental results from various scenarios show that the proposed two-phase algorithm effectively reduces Q-commerce delivery costs by approximately 10.97% compared with initial solutions. Of note, simultaneously considering synchronous and asynchronous transshipment resulted in an additional 0.5 percentage points (pp) improvement in the objective function value compared with using a single transshipment. These findings suggest that transshipment solutions can effectively reduce vehicle operating times and request travel distances, contributing to the future popularization of multi-echelon delivery systems.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709227</guid></item><item><title>A Data-Driven Simulation and Machine Learning Framework for Shopping Trip Forecasting with Spatial Clustering</title><link>http://pubsindex.trb.org/view/2709131</link><description><![CDATA[Retailing plays a pivotal role in the functioning of urban systems. While upstream supply chain activities such as manufacturing and distribution primarily affect freight movement, the retail interface translates consumer demand into individual travel behavior, shaping local traffic conditions and feeding back into upstream logistics. Despite its importance, shopping-related travel remains under-modeled in urban mobility research. To address this gap, this study develops a purpose-specific travel forecasting and simulation framework for predicting shopping trip demand in urban areas. The forecasting model integrates commercial-environment attributes, trip characteristics, and sociodemographic factors. A suite of machine learning (ML) models is evaluated, and the best-performing model is selected for the proposed simulation. Microlevel predictions are then scaled to the full urban region, followed by zonal aggregation and k-means spatial clustering to identify distinct retail-demand patterns and support scenario testing. Numerical results show that the random forest model outperforms alternative ML classifiers and, when implemented in the simulation, generates a citywide estimate indicating that shopping trips represent 14.3% of all weekday travel, in line with external regional benchmarks. The combined ML–simulation framework demonstrates strong predictive performance and reveals meaningful spatial and behavioral insights relevant to policymaking and planning applications. Although applied to Halifax, the modular structure of the framework makes it transferable to other urban regions and adaptable to additional trip purposes, supporting future extensions involving multiactivity modeling, causal impact analysis, and integration with passive mobility datasets.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709131</guid></item><item><title>Exploring Aggregate Morphological Characteristics under Laboratory Polishing for Enhanced Pavement Skid Resistance</title><link>http://pubsindex.trb.org/view/2709130</link><description><![CDATA[As the use of recycled asphalt pavement (RAP) in pavement construction grows for sustainable development, it becomes essential to investigate potential frictional deterioration over time. This study evaluated the friction properties of recovered RAP material aggregates compared with raw aggregates across various polishing cycles. The micro-Deval test was employed to simulate aggregate loss of texture, while morphological and friction properties were measured using an aggregate imaging measurement system (AIMS-II), along with a British pendulum tester (BPT) and dynamic friction tester (DFT). Additionally, Fourier transform infrared spectroscopy (FTIR) was employed to assess its potential in determining the origin and composition of RAP material aggregates. A simple method was used to fabricate custom aggregate rings, allowing for accurate testing in the DFT setup. The aggregate testing results revealed notable variations across the measurement techniques. AIMS-II analysis showed that traprock (maroon-colored) exhibited the highest surface texture, while DFT and BPT results indicated that certain limestones outperformed traprock in friction properties. Additionally, the testing results demonstrated that the RAP materials were comparable to, or even outperformed, certain limestone sources. However, because of potential variability within RAP stockpiles, careful quantification is necessary to assess their suitability. FTIR analysis demonstrated its ability to distinguish between carbonate-rich and silica-rich aggregates; however, further research is needed to build a library of aggregate sources. Finally, a machine learning algorithm identified the loss of aggregate DFT₂₀ values as the most significant aggregate property representing friction loss in asphalt mixtures.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709130</guid></item><item><title>Evaluating the Safety Impact of Roadway Rightsizing in Jefferson County, Kentucky</title><link>http://pubsindex.trb.org/view/2709128</link><description><![CDATA[This research provides a safety assessment of rightsizing projects that took place in Jefferson County, Kentucky. Rightsizing has become increasingly popular as a solution for multimodal access improvements and enhancing roadway safety. A cross-sectional before–after analysis was applied to a 15-year panel dataset from 2010 to 2024 to estimate the impact of rightsizing on crash frequency. A matched control group was developed using traffic volume and segment length using nearest-neighbor approach. Negative binomial safety performance functions were estimated with untreated sites and adjusted with annual calibration factors for seasonal changes consideration. Empirical Bayes methods were applied to correct for regression-to-the-mean bias and estimate counterfactual crash frequencies. Crash modification factors (CMFs) were calculated and disaggregated by crash type (all, bicycle, pedestrian, and intersection-related) and severity level (KA, BC, O). The analysis reveals that rightsizing treatments were associated with a 32% reduction in fatal and severe injury crashes, and consistent crash reductions at intersections. However, elevated CMFs across all severity levels for bicycle crashes suggest increased risk, potentially because of higher exposure without corresponding protective infrastructure. Pedestrian findings varied by severity level. The findings highlight crash severity reduction potential for rightsizing while indicating a requirement for including facilitative infrastructure for protection of vulnerable road users. The study includes practical recommendations for transportation agencies considering rightsizing as part of a broader safety and multimodal mobility initiative.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709128</guid></item><item><title>Factors Associated with the Gender Gap in the Mobility of Care within Two-Parent Households in Mexico City</title><link>http://pubsindex.trb.org/view/2708409</link><description><![CDATA[Mobility of care (MOC) is a term that has recently been introduced into the transportation planning literature to refer to travel related to activities needed for the upkeep of the home and the well-being of its members. Gender bias is an issue since such travel is mainly undertaken by women and remains largely hidden in traditional travel surveys. The emergent quantitative work has focused on identifying MOC patterns across individual aspects; however, this approach limits the understanding of the interdependent nature of MOC among household members. There is a lack of household-level studies which would help better understand the MOC gender gap. This study in the Mexico City Metropolitan Area develops robust regression models to understand the determinants of the MOC gender gap, controlling for two-parent household structures. The results consistently demonstrate that travel time to work is strongly and inversely associated with the availability to participate in MOC for the head-of-household (predominantly men) and the spouse (predominantly women). The spouse who is dedicated to home chores is the main actor carrying out MOC, increasing the gap in their larger load as the household grows (e.g., with children in basic education), and when MOC is done solely on weekdays. The built environment showed limited associations with the gender MOC gap. Other significant associations, and results were compared with the literature and discussed for their relevance to reducing the gender gap.]]></description><pubDate>Tue, 02 Jun 2026 11:01:49 GMT</pubDate><guid>http://pubsindex.trb.org/view/2708409</guid></item><item><title>Comparison of Pavement Roughness Indicators from Traditional Inertial Profilers and Emerging Connected Vehicle Sensors</title><link>http://pubsindex.trb.org/view/2709233</link><description><![CDATA[This study is an evaluation of the consistency and potential of crowdsourced connected vehicle (CV) data as an alternative to traditional inertial profiler (IP) measurements for road roughness evaluation. IP data were collected from three roadway corridors, SH 21, SH 6, and FM 2818, in Bryan, Texas, which comprised a variety of pavement types and functional classifications. Lane-specific international roughness index (IRI) values were recorded using a calibrated inertial profiler and analyzed at every 10 ft. These were compared with direction level ride values collected at an average spacing of 82 ft from a third-party CV data provider. A general agreement was found between IP and CV data on asphalt and concrete surfaces, but there were substantial differences between seal coat sections, where CV ride values overestimated roughness by 80–100 in./mi. A grouping analysis was conducted to compare segments categorized by roughness level from both datasets. The results showed a moderate match in identifying the roughest segments by the CV system, compared with those identified by IP. This match rate varied by profile and improved with broader grouping sizes. Results show that CV data might have greater sensitivity to vehicle behavior (for example, braking at intersections), while IP-based systems are calibrated to record actual road roughness. This could be one of the reasons for the moderate mismatch in identifying rougher segments using the two methods. Despite these limitations, the CV system’s high frequency of measurements, especially on highways, demonstrates its potential for cost-effective, network-level pavement monitoring.]]></description><pubDate>Mon, 01 Jun 2026 16:52:46 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709233</guid></item><item><title>Optimization of Emergency Bus Bridging Service with Mixed Diesel–Electric Fleets under Metro Disruptions</title><link>http://pubsindex.trb.org/view/2709234</link><description><![CDATA[Metro service disruptions occur frequently in large cities, making emergency bus bridging services essential for maintaining network functionality and reducing passenger delays. With the rapid electrification of urban bus fleets, traditional bus bridging strategies face new challenges related to electric bus state-of-charge (SOC) limits and charging availability, which are often ignored in existing models. This study develops a mixed-fleet optimization framework for emergency bus bridging services that jointly considers diesel buses and electric buses (EBs). Two models are formulated: a standard feeder model (SFM) providing all-stop services, and a combined feeder model (CFM) that supplements standard feeders with limited-stop direct services to serve high-demand origin–destination (OD) pairs. Both models explicitly incorporate EB SOC constraints, charging requirements, and coordinated fleet deployment. A case study based on the December 2023 rear-end collision on the Beijing Subway Changping Line demonstrates that both models can generate feasible emergency response plans under realistic operational and charging constraints. Compared with the SFM, the CFM reduces total system cost from RMB 76,802.99 to 76,752.93. This improvement is driven by a reduction of RMB 322.98 in passenger delay cost, which outweighs an increase of RMB 272.93 in operator cost. Sensitivity analyses are further conducted to evaluate model performance under varying demand patterns, disruption durations, direct-service capacities, and charger availabilities. Results indicate that the CFM performs best under concentrated demand, long-distance trips, and moderate disruption durations, highlighting its effectiveness in reducing passenger delays for high-impact OD flows.]]></description><pubDate>Mon, 01 Jun 2026 16:52:46 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709234</guid></item><item><title>Cold-in-Place Recycling with 100% Recycled Asphalt Pavement Rejuvenated by Soybean Oil: Laboratory and Field Evaluation</title><link>http://pubsindex.trb.org/view/2709232</link><description><![CDATA[This study investigated the feasibility and performance of using soybean oil as a bio-based recycling agent in recycled asphalt pavement (RAP) for road reconstruction in cold regions. A comprehensive demonstration project was conducted on a 5-mile section of Old State Road in Clare County, Michigan, where a 100% RAP mixture modified with soybean oil was produced and placed using a conventional asphalt paver equipped with a screed. Laboratory evaluations included balanced mix design, rutting and cracking testing, and binder performance analysis. Field application processes, including mixing and compaction, were also documented and evaluated. The asphalt mixture tests included the Hamburg wheel-tracking test (HWTT) and the indirect tensile asphalt cracking test (IDEAL-CT), while the asphalt binder tests included dynamic shear rheometer (DSR), asphalt binder cracking device (ABCD), rotational viscometer (RV), linear amplitude sweep (LAS), Fourier transform infrared spectroscopy (FTIR), and CO2 emission analysis. An optimal soybean oil dosage of 1.0 wt.% (based on the total weight of the mix) significantly improved low-temperature cracking resistance and fatigue life while maintaining rutting resistance. Results showed that soybean oil improved compaction performance and exhibited a cracking temperature approximately 3.3°C lower than that of untreated RAP based on the ABCD test. Fatigue performance was also enhanced. Fourier transform infrared spectroscopy (FTIR) analysis confirmed the chemical compatibility and interaction between soybean oil and the RAP binder. On-site application was completed smoothly without workability issues, and the final pavement met all compaction and density requirements. In summary, using soybean oil as an RAP recycling agent provides a practical and environmentally friendly solution to improve the performance of recycled asphalt mixes, especially for low-volume roads in cold climates, while supporting the sustainability of Michigan’s pavement and the growth of the soybean market.]]></description><pubDate>Mon, 01 Jun 2026 16:52:46 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709232</guid></item><item><title>Evaluation of Critical Pavement Responses from Accelerated Pavement Testing on Airfield Flexible Pavements Surfaced with Hot and Warm Mix Asphalt</title><link>http://pubsindex.trb.org/view/2709231</link><description><![CDATA[Highway agencies frequently use warm mix additives as compaction aids. Lower production temperature of the warm mixes simultaneously entails the benefit of widening the paving window. Airport authorities can ensure significant fiscal savings with reduced downtime through the adoption of similar technologies in airfield paving. However, limited scientific information exists concerning the performance of these materials in airside flexible pavements. Aircraft gross weights and tire pressures have also been routinely increasing over the last few decades with the advent of new-generation aircraft. The Federal Aviation Administration (FAA) procured a sixth-generation heavy vehicle simulator, airfields (HVS-A) to investigate the performances of resilient pavement materials under simulated aircraft loading. Accordingly, six full-scale test lanes were constructed during Test Cycle 1 (TC1) at FAA’s National Airport Pavement and Materials Research Center (NAPMRC) using four different asphalt concrete (AC) mixes with two different binder grades. Each test lane was divided into three test sections. Asphalt strain gauges and pressure cells were installed in the test sections to monitor the critical pavement responses over the duration of traffic tests. Corresponding test sections were trafficked under different combinations of high tire pressure and temperature. This paper examines the tensile strains at the bottom of AC and compressive stresses on top of the subgrade in reference to the observed rutting performances in four TC1 outdoor test lanes. The respective hot and warm mixes exhibited comparable rutting performances, and the sensor observations corroborated the related findings.]]></description><pubDate>Mon, 01 Jun 2026 16:52:46 GMT</pubDate><guid>http://pubsindex.trb.org/view/2709231</guid></item><item><title>Examining the Role of the Built Environment in Cycling Injury Severity: Older Adults (60+) Versus Individuals Aged 10 to 59 in a Super-Aged City</title><link>http://pubsindex.trb.org/view/2706341</link><description><![CDATA[Cycling offers well-documented benefits, including reduced congestion and air pollution, enhanced mobility, and improved physical health. Reflecting these advantages, cycling participation has increased across all age groups in many developed countries. However, this growth has been accompanied by a rise in cycling crashes, raising significant urban safety concerns—particularly for older adults. Although numerous studies have investigated factors influencing the injury severity of cycling crashes, the built environment has consistently emerged as a key determinant. Nevertheless, limited research has specifically explored how micro-level built-environment characteristics are associated with the injury severity of bicycle crashes, especially among older adults. This study investigates the association between built-environment characteristics and the injury severity of bicycle crashes involving older adults, analyzing 10,502 crash cases in Seoul from 2018 to 2023 using a binomial logistic regression model. To capture detailed built-environment attributes, we applied DeepLabV3+ for semantic segmentation of Google Street View images collected from four directions at each crash location. The results indicated that higher proportions of road surfaces, obstacles, and vegetation were associated with increased injury severity among older adults (60+), whereas the presence of traffic devices reduced injury severity. Among individuals aged 10 to 59, greater building density was linked to lower injury severity. A common risk factor across both age groups was collisions with motor vehicles. These findings underscore the necessity of age-sensitive safety interventions. For older adults in particular, measures such as separating cycling paths from obstacles and increasing the installation of traffic control devices may help improve cycling safety.]]></description><pubDate>Thu, 28 May 2026 10:47:37 GMT</pubDate><guid>http://pubsindex.trb.org/view/2706341</guid></item><item><title>Safety Risks of Out-of-Context Curves: Three Decades of Rural Curve Research in New Zealand</title><link>http://pubsindex.trb.org/view/2706340</link><description><![CDATA[The rural road network of New Zealand contains many horizontal curves that are inconsistent with their surrounding environment. These “out of context” curves—where the safe negotiation speed is significantly lower than the prevailing approach speed—are associated with higher crash risk than in-context curves of otherwise similar geometry. Over the past three decades, New Zealand researchers and transport agencies have developed a robust body of work to understand and address this issue, although it remains underexplored internationally. This paper reviews the evolution of rural curve safety research in New Zealand, including the development of high-resolution road geometry datasets, operating speed models, and crash prediction models. It also highlights how these insights have informed national project evaluation guidance and safety prioritization frameworks, and a recent adaptation of this research to international contexts, including the United States. In particular, it was found that traditional crash prediction models such as those in the Interactive Highway Safety Design Model and the Highway Safety Manual can underestimate the observed number of crashes around out-of-context curves by at least 30%, and potentially up to 60%. The recent application of New Zealand curve context modeling to U.S. rural roads through the SafeCurves software tool addresses this limitation. This review aims to demonstrate the value of incorporating curve context into safety analysis and prioritization, highlight New Zealand research in this area, and encourage broader application of these methods to reduce crash risk on rural roads worldwide.]]></description><pubDate>Thu, 28 May 2026 10:47:37 GMT</pubDate><guid>http://pubsindex.trb.org/view/2706340</guid></item><item><title>Toward Asphalt Pavement Construction Safety Improvement with Generative Artificial Intelligence</title><link>http://pubsindex.trb.org/view/2706339</link><description><![CDATA[Recent advancements in generative artificial intelligence (GenAI), particularly large language models (LLMs), have shown promise in enhancing safety analysis within the construction industry. This study explores the integration of structured information from the U.S. Pennsylvania Department of Transportation’s job safety analysis (JSA) documents with unstructured accident narratives from the U.S. Occupational Safety and Health Administration’s Integrated Management Information System (IMIS), focusing on asphalt pavement construction—a sector marked by complex operations and hazardous equipment. Multiple LLMs were employed to classify accident narratives across four dimensions: construction type, relevant JSA job and step, environmental or operational influence, and hazard type. This approach aims to identify frequently cited work activities, assess gaps in safety documentation, and improve future hazard recognition. While general job classification achieved moderate success, performance declined for step-level and contextual classifications, largely because of ambiguous language and overlapping job responsibilities in the narratives. Despite these limitations, LLMs uncovered critical patterns. Paver and roller operations emerged as high-risk activities, often influenced by environmental factors such as traffic or weather. Furthermore, exploratory hazard analysis revealed that model-suggested hazard labels were sometimes more contextually appropriate than those in the original IMIS database, indicating opportunities for data refinement. By aligning structured safety documentation with real-world incidents through GenAI, this study highlights a novel pathway for data-driven safety planning in highway construction. While expert oversight remains essential, the results demonstrate the potential of LLMs to support more adaptive, targeted, and proactive approaches to risk assessment and mitigation.]]></description><pubDate>Thu, 28 May 2026 10:47:37 GMT</pubDate><guid>http://pubsindex.trb.org/view/2706339</guid></item></channel></rss>