RPUG 2026 Speakers’ Bios and Abstracts

Abstract:
This event includes two transportation pooled fund (TPF) meetings:
• TPF-5(537) Improving the Quality of Highway Profile Measurements, and
• TPF-5(570) Improve Pavement Surface Distress and Transverse Profile Data Collection and Analysis, Phase III

These meetings are open to their TPF members and all others interested in these subjects.

Abstract:
It is an icebreaker reception and a great networking opportunity to be held in the exhibit hall.

All RPUG attendees are welcome!

Abstract:

Abstract:

Abstract:

Abstract:

Abstract:
This presentation addresses the use of IRI as a measure of functional pavement quality on urban and low-speed roadways. The interaction of travel speed, vehicle properties, and the roughness content on urban and low-speed roadways is discussed. In particular, vehicles are sensitive to different road features are lower speed than they are at higher speed, and the relationship between the IRI and the intensity of vehicle response changes with speed.
The presentation describes an experiment that was conducted to correlate measured road roughness to objective measurements of vibration experienced by a vehicle driver. Three instrumented vehicles were tested on 29 urban and low-speed test sections. The testing included passes over each test section at speeds in the range from 25 mi/hr to 45 mi/hr. The instrumentation provided simultaneous measurements of road profile and acceleration at several interfaces between the vehicle and the driver.
The experimental results produced two observations with practical implications. First, the vibration experienced by the driver on the majority of the test sections included content that classified as “transient” because of the presence of localized roughness. Second, for travel speeds down to 25 mi/hr, adjustment of the IRI algorithm toward lower simulated travel speed improved correlation to measured discomfort experienced by the driver.
The presentation offers suggestions for using the modified versions of the IRI algorithm to quantify roughness on urban and low-speed roadways, and discusses their strengths and weaknesses.
The presentation is based on work published in NCHRP Report 914.

Abstract:
The use of contractor data for quality assurance requires an independent assurance program. The Texas Department of Transportation (TxDOT) uses contractor measurements to make bonus and penalty payments for ride quality. TxDOT’s independent assurance ride quality verification program is now over 10 years old. Lessons have been learned on project tracking, verification rates, verification results, and practical data collection practices. This study shows that TxDOT’s verification program meets or exceeds the testing frequency required in the Code of Federal Regulations, and discusses the nuances associated with calculating the verification frequency. This study also shows that the average difference between the contractor’s data and verification data has been below the 3 in./mi. (0.05 m/km) threshold to accept the contractor’s data in each fiscal year. For the 411 projects verified between 2015 and 2024, the contractor’s data have been acceptable 92 percent of the time. For the remaining 8 percent, TxDOT could elect to use the contractors’ data, verification data, or an average of the two. Referee testing has not been required on any verification project. This study also discusses the importance of timely data submission and potential struggles with having a verification clock attached to the contractors’ data collection instead of the contractors’ data submission. This has proven especially difficult for concrete projects, further challenged by their duration. State departments of transportation can use this study to help develop or improve their ride quality verification program. This study briefly highlights the additional benefit of creating a subset of ride quality data on construction projects that can be tracked into the future.

Abstract:
Pavement roughness is a key indicator of pavement functional performance, with direct implications for user comfort, vehicle operating costs, safety, and long-term pavement preservation. Smoother pavements at the time of construction have been shown to delay the onset of maintenance and rehabilitation needs, while pavements that enter service with higher roughness levels may experience accelerated deterioration and increased vehicle fuel consumption due to amplified dynamic loading from traffic. In response to these impacts, state transportation agencies across the United States have adopted rideability specifications that define acceptable smoothness levels for newly constructed and rehabilitated pavements. The South Carolina Department of Transportation (SCDOT) has implemented a comprehensive rideability specification based on the International Roughness Index (IRI), with acceptance thresholds tailored to different project types, such as new construction, pavement overlays, and others. While the current specification reflects established practices and project-specific considerations, a statewide, project-level evaluation of historical ride quality outcomes has not been conducted to assess long-term trends, practical achievability, and the distribution of achieved IRI values relative to the specified acceptance limits. This study presents a statistical analysis of SCDOT’s historical project-level rideability data collected over the past decade. The analysis examines temporal trends in ride quality outcomes, evaluates the frequency with which incentive and disincentive thresholds are met, and characterizes the distribution of achieved IRI values across project types. Based on these findings, a statistically based framework using cumulative probability functions is introduced to support data-informed evaluation of and potential adjustments to the rideability specifications. The results provide insight into the reasonableness and achievability of current IRI thresholds and offer a basis for exploring preliminary adjustments to rideability limits for different project categories.

Abstract:
This presentation describes upgrades to the Federal Highway Administration Benchmark Profiler (BP), the development of a High-Speed Reference Profiler (HSRP), and comparison between profiles measured by the two devices.
In 2024, testing at the Illinois Certification and Research Track (ICART) demonstrated agreement between the BP and the University of Michigan Transportation Research Institute Urban and Low-Speed Profiler (ULSP) above the threshold established for reference quality profilers. Since then, the BP has been upgraded, and the HSRP has been developed as a less complex version of the ULSP. The upgraded BP will be deployed at ICART and a ground-truth measurement device, and the HSRP will be deployed at ICART as an efficient device for measuring reference profiles frequently in support of the profiler certification program.
BP, ULSP, and HSRP measurements were compared on six pavement sections at ICART. The test sections composed a diverse set of macrotexture types. Agreement between profiles was quantified using cross correlation of profile filtered by the International Roughness Index algorithm.

Abstract:
Traditional inertial response–based profilers have long supported pavement management with longitudinal measurements for roughness and surface level defects, but their 2D focus limits characterization of patching, material transitions, localized discontinuities, and nonuniform repairs. This paper presents an INS integrated, multimodal roadway profiling architecture—implemented within Fugro’s ARAN® Elite platform—that produces spatially coherent, 3D datasets combining high-rate pavement profiling with 360° oblique LiDAR under a unified inertial reference frame.

In a single pass, the system delivers a full lane, high-resolution 3D surface map with dense sampling that supports generating profiles wherever needed within the lane. This enables analyses beyond classical profilers, including identification of surface expressed patch boundaries, detection of vertical discontinuities at the surface, recognition of material transitions evident in surface geometry, and examination of macrolevel condition patterns associated with drainage and shoulder characteristics.
The resulting multimodal datasets support a wide range of research pathways, including machine learning methods for surface condition inference, studies of corridor level influences on profile shape, and enhanced roadway condition modeling that leverages correlated surface and 3D context data. The system aligns with established profiling standards (AASHTO R 56; ASTM E950/E1926) and yields results which pass NCAT Class I Profiler certification.

To help advance the state of practice, Fugro will make representative, spatially coherent multimodal datasets available to qualified researchers through collaborative data use arrangements, enabling non-commercial exploration of additional insights that these rich datasets may unlock — including multimodal fusion analytics, advanced surface condition characterization, and new approaches to understanding corridor level influences on pavement performance.

Abstract:
Accurate pavement distress segmentation is critical for roadway maintenance and traffic safety. However, current segmentation models often struggle with underrepresented classes due to data imbalance and the high cost of collecting and annotating diverse, high‑quality images. This study introduces PaveFormer, a multi‑modal segmentation framework that integrates class‑specific, diffusion‑based data augmentation with digital image processing to improve segmentation accuracy and generalization. The framework adapts Stable Diffusion v1.5 through transfer learning and a ControlNet‑based conditioning strategy, employing a multi‑stage design to generate high‑fidelity images. A customized loss function increases emphasis on hard‑to‑learn samples, and a Top‑k filtering approach selects synthetic images based on crack and pothole pixel counts to ensure augmentation quality. Using 5,221 original labeled images, the system produces 46,061 training samples covering eight pavement distress categories: background, crack, pothole, sealed crack, manhole, lane marking, expansion joint, and patch. PaveFormer, implemented with low‑rank adaptation fine‑tuning and targeted architectural enhancements to the pre‑trained SegFormer‑B5 model, demonstrates strong performance through model comparisons, ablation studies, and cross‑dataset evaluations. The full framework is available on GitHub.

Abstract:
Recent advances in artificial intelligence (AI) and imagery-based analysis have expanded the ability to detect and classify pavement surface distresses at scale. While AI-based defect detection has demonstrated effectiveness at the project level, challenges remain in applying these methods consistently across large, distributed pavement networks while maintaining data quality, repeatability, and usefulness for pavement surface characterization (PSC).

This presentation examines the application of AI-assisted pavement distress detection integrated with GIS frameworks to support network-level pavement surface measurements. The focus is on how automated distress outputs—derived from vehicle-based and aerial imagery—can be standardized, validated, and spatially organized to support consistent PSC reporting across multiple facility types and survey cycles.

In addition to measurement considerations, the presentation emphasizes structuring PSC data for downstream decision-making. Automated distress outputs are evaluated not only for technical accuracy, but for their ability to support maintenance scoping, cost development, and timely preparation of construction-ready repair packages. This application-oriented perspective highlights the importance of aligning data resolution, classification logic, and spatial accuracy with real-world implementation needs.

Key topics include scaling AI-based distress detection across large pavement inventories, integrating AI-derived surface distress data into GIS-based network models, managing variability due to surface texture and collection conditions, and translating defect-level detections into actionable, network-level condition indicators.

Rather than emphasizing specific algorithms or proprietary tools, the presentation focuses on process design, data governance, and output usability, highlighting lessons learned from multi-year application of AI-assisted PSC across large pavement networks.

Abstract:
Road agencies need to maintain an up-to-date database of maintenance treatments performed on their network for asset management, forecasting, and planning. One common repair activity for jointed plain concrete pavement (JPCP) is to replace cracked concrete slabs with new ones. Images of the pavement surface are collected as part of the Automated Pavement Condition Survey (APCS); however, these images currently need to be reviewed and manually labeled by an operator as a replaced slab, which is time-consuming, and inconsistent. To address this issue, this study presents a novel deep learning (DL) approach for automated detection of replaced slabs using high-resolution 3D pavement surface images. While prior research has focused primarily on surface-level distress detection, such as cracking, this work is the first to apply deep learning methods to the problem of identifying replaced slabs. This fills a critical gap in pavement asset management research. We developed a sequence-aware model that combines a TimeDistributed Convolutional Neural Network (CNN) for analyzing the image with a Gated Recurrent Unit (GRU) network (which processes sequences of slabs) to better understand how slabs appear across space. In addition to raw image data, we also added engineered features (intensity histograms and Local Binary Pattern (LBP) histograms). They are computed from the green channel. All slab images were uniformly cropped to 400×1536 pixels using a full-width, bottom-cropping strategy to standardize the inputs. The model was trained on about 10,000 labeled slab images. It reached 80.7% accuracy, with 75.7% precision and 86.4% recall on the test data. Errors happened on slabs that looked ambiguous even to humans. Still, the model showed strong potential to replace manual inspection with an automated and scalable tool. This work shows that deep learning can be used to detect replaced slabs and opens the door for future use on more realistic pavement datasets.

Abstract:
Automated pavement condition assessment increasingly relies on both modalities of 2D intensity and 3D surface imagery collected at highway speeds. Converting these imagery data into accurate, reliable, and robust machine learning models for automated pavement distress detection and segmentation, requires pixel-level segmentation masks that preserve distress geometrical information. A practical barrier is annotation: high-performing segmentation models require large, consistent mask datasets, yet the labeling is time consuming and error-prone, especially for thin cracks and joint-adjacent distress, and outcomes vary across workflows. This study introduces a cost-effective pipeline that targets (1) a faster, more repeatable mask annotation method, and (2) performance-based selection of segmentation models that balance mask fidelity and inference speed for deployment. A standardized labeling protocol and the repeatability under manual and assisted refinement are verified. We also test simple, local contrast sharpening inside each bounding-box region to make thin cracks easier to see during assisted labeling. Using a multimodality dataset spanning asphalt and concrete pavement, a transformer-based model (e.g., SegFormer-B2) and a real-time YOLO-based model (e.g., YOLOv8-seg) are compared. Initial research findings indicate that structured procedures with quality control reduce rework and improve thin-crack consistency, while multimodal inputs improve continuity in low-contrast regions. This research provides a practical solution to cost-effective imagery annotation and dataset preparation and will lead to more rapid development of machine learning models to support automated distress detection and segmentation for pavement asset management.

Abstract:
Hydroplaning is a critical wet-weather safety phenomenon governed by the interaction between vehicle dynamics, environmental conditions, and pavement surface characteristics. Crash narratives provide detailed qualitative descriptions of roadway and environmental conditions and are commonly used to identify hydroplaning-related crashes. Pavement surface macrotexture quantified through road profile measurements, such as Mean Profile Depth (MPD), directly influences pavement surface drainage capacity and water film formation and therefore constitutes an important engineering parameter in hydroplaning analyses. Despite their complementary relevance, crash narratives and measured pavement surface texture data are seldom integrated within a unified systematic framework. Considering this gap, this study proposes a data-driven approach that integrates crash narrative-based hydroplaning identification with pavement surface texture metrics to enhance the interpretation of hydroplaning-related crash occurrences in asphalt pavements.

The proposed analysis focuses on the standard crash data retrieved from the Crash Report Information System for the Austin district for FY2020-2024, comprising crash narratives, roadway attributes, and geospatial information. Hydroplaning-related crashes will be identified through a structured keyword-based filtering of narrative text, capturing wet pavement surfaces and vehicle loss-of-control descriptors. Further, crash records will be spatially and linearly matched to pavement sections using latitude/longitude coordinates and distance-from-origin information, enabling the assignment of corresponding MPD metrics to individual crash events. Comparative statistical analyses will be conducted to evaluate differences in pavement macrotexture characteristics between narrative-identified hydroplaning and non-hydroplaning crashes.

The integrated narrative-texture analysis is expected to provide quantitative insight into the relationship between pavement macrotexture and hydroplaning-related crash occurrence, while preserving the contextual value of narrative descriptions. The proposed framework supports engineering-informed evaluation of wet-weather crash patterns and illustrates the practical value of incorporating road profile measurements into road safety analyses. Overall, the study findings will support network-level hydroplaning risk assessment and texture-related pavement maintenance decision-making through integrated analysis of crash narratives and road profile data.

Abstract:
This research aims to automate the time-consuming but essential task of labeling pavement distresses with precise segmentation masks. The proposed approach fully utilizes existing 2D/3D image datasets with bounding boxes, applying the Segment Anything Model 3 (SAM3) to localize and generate initial distress masks within a local Computer Vision Annotation Tool (CVAT). Human experts refine these masks with significantly less effort. We will report a timing study comparing manual from-scratch segmentation versus box-guided SAM3 plus refinement. Additionally, the problem of rare distress classes is addressed through this solution since more specific masks are provided. We will initially report results on failed concrete patches and popouts, the two poorest-performing classes in our box-only baseline, as a stress test of the proposed box-to-mask pipeline. Using the resulting segmentation labels without increasing distress instances, a lightweight YOLO model improves mAP50 from 0.464 and 0.007 to 0.761 and 0.69, respectively; the same pipeline generalizes to other distress types. Furthermore, the newly trained segmentation model will be integrated back into CVAT, to create a feedback loop for more accurate annotations. This work could meet practical needs of research community by reducing data leakage risk and allowing project-specific customization. This research work may significantly improve the time cost efficiency and data quality in the preparation of segmentation annotation of pavement images for the development of AI models for automated pavement distress detection and segmentation.

Abstract:
Accurate and timely identification of pavement raveling (loss of aggregate) is critical for roadway safety, ride quality, and cost-effective pavement preservation. Although machine learning (ML)–based raveling detection models have been successfully developed in prior research, including NCHRP IDEA projects, implementing these models in operational environments remains challenging for many state Departments of Transportation (DOTs). Challenges arise not from model accuracy alone, but from data readiness, annotation practices, quality assurance, and organizational capacity for adopting artificial intelligence (AI) tools.

This study presents implementation experiences of automatic raveling classification using ML, with a focus on pilot applications conducted in collaboration with FDOT and MDOT. Rather than proposing new algorithms, the paper emphasizes a system-level, life-cycle approach to AI-enabled problem solving, covering data selection, annotation and calibration, model training or deployment, QA/QC, result interpretation, and integration with pavement management workflows.

Through these implementation efforts, several key findings emerged. First, DOTs differ substantially in their ability to adopt ML at different levels: agencies with in-house AI expertise can implement and retrain models directly, while others achieve more reliable outcomes by adopting application-level tools with pre-trained models and structured QA/QC processes. Second, annotation quality and consistency were found to be the dominant factors affecting implementation success, often outweighing the choice of ML model itself. Calibration among raters, clear severity definitions, and balanced datasets are essential for producing stable and interpretable results. Third, domain differences in sensing platforms, pavement textures, and mixed distress conditions necessitate explicit QA/QC procedures and, in some cases, targeted retraining before results can be trusted for decision-making.

Overall, this work highlights practical lessons learned from transitioning ML-based raveling detection from research to practice and provides guidance for DOTs seeking to adopt AI technologies in a scalable and operationally meaningful manner.

Abstract:
Pavement Management Systems (PMS) rely on accurate and consistent pavement condition data to support effective maintenance and rehabilitation (M&R) planning. Typical pavement condition data includes: Roughness Data, commonly quantified as International Roughness Index (IRI) and obtained using a full-inertial profiler; Rut Data, expressed as the depth of pavement surface depression in the wheel path, and, Surface Distress Data, rated in terms of distress type, severity, and quantity in accordance with certain standards and converted into a composite distress index. Traditionally, surface distress assessments were based on windshield surveys done by raters in a survey vehicle. As imaging and pattern recognition technologies have evolved, transportation agencies have been moving away towards automating the collection and compilation of crack/distress data. The Laser Crack Measurement System (LCMS) is one such technology – a 3D-laser system – capable of obtaining automated cracks, rutting and other surface distresses. This study evaluates and compares three (3) years of pavement condition data collected on a fifty-seven (57) mile segment of the New Jersey Turnpike. LCMS imagery and inertial profiler data were obtained on the most traveled lane. Also, an in-house developed algorithm was used to categorize and quantify crack type, severity, and percentage automatically according to the New Jersey Department of Transportation’s (NJDOT’s) Surface Distress Index (SDI) distress definitions and the resulting SDI values were calculated. The SDI uses a 0-5 scale and is akin to the Federal Highway Administration’s (FHWA’s) Present Serviceability Rating (PSR). Overall, it was found that IRI slightly increased and SDI decreased on pavement sections where no M&R was done over the 3-year data collection period. The SDI index was more sensitive to identify pavement deterioration over time. Based on these findings, the LCMS and complementary in-house developed algorithm proved to be successful tools for the automated collection and analysis of network pavement condition.

Abstract:
Traditional accelerometer‑based road profiling methods often perform poorly in urban environments due to stop‑and‑go traffic, which introduces nonlinear noise that standard filtering techniques cannot effectively separate from true pavement geometry. This study develops and validates a transformer‑based deep learning model designed to reconstruct longitudinal road profiles under these challenging conditions. The model integrates linear acceleration and angular velocity from an Inertial Measurement Unit (IMU) with Global Positioning System (GPS) spatial positioning and is trained using datasets collected from multiple urban road segments, with ground‑truth profiles provided by a SurPRO walking profiler. The transformer architecture employs self‑attention to process full profiling sequences and iteratively minimizes residual errors between predicted and reference profiles, enabling separation of pavement features from vehicle dynamics. A loop‑closure elevation check is used to evaluate the model’s ability to control vertical drift. Results demonstrate that the proposed method can accurately reconstruct road profiles and produce reliable IRI measurements despite frequent acceleration and deceleration events in urban conditions.

Abstract:
In the evolving landscape of construction and engineering, precision is paramount. Explore the benefits of using various geospatial solutions and selecting the proper tools for the application. “”Geospatial Positioning Based Technologies to Improve Road Quality and Smoothness “” explains the various geolocation abilities and technologies and how they are incorporated into partnering technologies to improve safety, production, longevity and overall roadway quality. The transformative power of GNSS, Ground Based Survey systems, Networks, and LiDAR technology to enhance AMG and Data Collection, is paramount to a successful and long lasting pavement. This presentation delves into how these technologies enables accurate modeling and control that ultimately improves project outcomes. Attendees will discover innovative applications in 3D machine control, and data collection from optimizing grading processes to enhancing safety and efficiency on-site.

Abstract:
Pavement skid resistance is a critical functional characteristic that controls vehicle maneuvers, particularly under wet weather, when surface water reduces available tire-pavement interface grip. While previous studies have primarily focused on categorizing skid resistance demand, the impact of available macrotexture is often overlooked. This demand-focused perspective limits the effectiveness of the relationship between skid resistance demand and safety. To address this gap, the present study introduces a novel network-level assessment that integrates skid-resistance demand and available pavement macrotexture to determine equilibrium.

Network-level friction and texture data were measured across North Carolina’s primary road network using the Sideway-Force Coefficient Routine Investigation Machine (SCRIM). Skid resistance demand was categorized into low, moderate, and high levels by integrating highway alignment (radius of curvature), highway features/environment (operating and hydroplaning speed difference), traffic characteristics (AADT), and interchange presence. Available macrotexture, measured as mean profile depth (MPD), was similarly categorized. The resulting skid resistance demand–available macrotexture framework produced a 3×3 matrix defining nine distinct zones. This framework was validated using a normalized wet lane-departure crash frequency (crashes per section length). Although crashes are rare events and typically require 3–5 years for analysis, pavement friction and macrotexture are time-dependent surface conditions that may vary on shorter time scales. Accordingly, crash counts were aggregated over a 13-month window centered on the surface measurement month, including 6 months before and 6 months after the measurement.

Results indicate that normalized crash frequency increases as skid resistance demand increases and macrotexture decreases. Sections characterized by high skid resistance demand and low macrotexture exhibited the highest median normalized crash frequency, whereas sections with low skid resistance demand and higher macrotexture showed the lowest. This proposed framework offers an easy and interpretable approach to network-level friction demand and available macrotexture to identify locations that warrant further investigation.

Abstract:
Transportation agencies have historically struggled with providing accurate crack length and quantity estimates for preventative maintenance crack seal projects. This is due to inconsistent measurement practices and reliance on manual, field-based methods. This project introduces an automated crack length estimation tool developed by the Montana Department of Transportation that leverages existing surface distress imagery and datasets to output standardized crack seal estimates.
The tool is built upon Pathway Services’ AutoCrack, a machine learning-based distress detection software that processes statewide 3D surface imagery and laser scans to identify and measure pavement cracks. To evaluate accuracy, a stratified random sample of 70 tenth-mile sections were selected across varying systems (Interstate, Primary, Secondary, Urban). Ground truth measurements were manually collected using surface imagery, right-of-way cameras, and LiDAR scans to account for adjustments where cracks veered out of range of the surface images. AutoCrack outputs were then compared to these measurements, and model parameters were refined to improve crack length estimation.
After adjustments were made the tool achieved a Mean Absolute Percentage Error of 17.9% and a crack identification accuracy of 92%. This represents a substantial improvement over traditional crack seal estimation methods, which have historically exhibited variability exceeding 50%.
This project demonstrates the feasibility of using automated pavement distress data to generate consistent and repeatable crack seal quantity estimates. Transitioning from manual estimation methods to an automated tool allows our maintenance crews to gather consistent, more accurate and efficient data that will improve over time with consistent use.

Abstract:
I hate speaking to large groups and generally think that vendors should stay off the podium at RPUG. However, sometimes vendors can offer a unique perspective on industry-level challenges. 25 years and 5 million miles of network-level rutting data in more than 40 states have taught me that the rutting statistic is far more complicated and challenging to get “right” than people think–and standardization is woefully overdue so that we can get DOTs and their equipment & data providers on the same page to measure and report the same statistic. I’d like to take a brand-agnostic trip through these very real challenges (theres a lot of reasons why it’s hard) and attempt to educate the industry on the critical variables in play, while pointing toward steps for a path to unify our efforts through research and discussion. In addition to leveraging real-world data and challenges, I plan to reach out to other equipment providers, educators and researchers to ensure the discussion is as representative as possible to the challenges facing us all. No brands, no sales, just technology, data and a call for industry-wide collaboration.

Abstract:
The Federal Highway Administration (FHWA) and the American Association of State Highway and Transportation Officials (AASHTO) have established standard practices for assessing transverse pavement profiler (TPP) measurements. As state departments of transportation (DOTs) move toward adopting these specifications, practical challenges related to testing procedures, data processing, and the consistency of test results have become increasingly evident. This presentation summarizes lessons learned from the analysis of TPP test data collected during recent demonstration projects conducted in Texas.
These projects involved controlled testing of multiple TPP systems. Reviews of the collected data identified limitations in existing TPP data analysis software, resulting in inconsistencies in calculation results and pass/fail outcomes. These issues increased data processing time and introduced variability that could adversely affect verification and acceptance decisions under future TPP-based profiler certification programs.
In response, subsequent demonstration projects at the Texas A&M Transportation Institute (TTI) focused on software enhancements and exploring potential refinements of existing pass/fail criteria. Improvements included automated data processing, algorithm refinements to produce more stable and reasonable results, and improved consistency in analysis outputs. These modifications substantially reduced data processing time and improved the repeatability and consistency of results across profiler systems and test runs.
The lessons learned highlight the importance of robust, transparent, and automated analysis tools for the reliable implementation of TPP specifications. The findings provide practical insights for agencies and practitioners preparing for TPP-based profiler certification and informed implementation and acceptance decisions.

Abstract:
Measurement of pavement smoothness has evolved from dragging wooden frames (Brown’s via graph) longitudinally along the pavement in the late 1800s to the use of inertial profilers and lidar systems today. Initially, two forms of evaluating roughness evolved: the measurement of the roadway profile and the measurement of a vehicle response to the traveled profile. Both approaches evolved independently and each had its advantages and disadvantages.
For acceptance of new concrete pavement construction, use of the California profilograph became the most widely used initial smoothness measurement device (1960) due to its light weight and convenience. Response type systems were better suited for existing roadway smoothness. The measurement of initial smoothness evolved around the operation of this device to a large extent. That is, measurement of one wheel path at a time and only one measurement per wheel path. With the development of the GM profilometer (1964) and the international roughness index (early 1980s), it was possible to relate accurate profile measurement with vehicle response.
The Profile Viewer and Analyzer (ProVAL) Software was initially developed by the FHWA to provide a means to view and analyze pavement profiles efficiently and robustly (2001). Several years after its introduction, ProVAL included a Grinding Simulation Model to determine, among other things, where best to diamond grind pavements to improve smoothness.
Although a major milestone at the time, the grinding simulation model did not properly represent commercial diamond grinding equipment operation. This presentation will address those issues, including how diamond grinding contractors use the software. An industry survey of practice has been conducted and will be presented. In addition, the fundamental characteristics of how a grinder works, including where profile is measured and where grinder footprints exist within the roadway, will be discussed.
Additionally, comments on reconsidering how to measure smoothness will be presented.

Abstract:
Persistent challenges arise when assessing pavement smoothness at low speeds or under stop-and-go conditions. These challenges compromise the accuracy of data collected by conventional profilers. For example, collecting ride quality measurements on a congested roadway with multiple driveways can lead to acceleration and deceleration of the collection vehicle, potentially adding “false” roughness into the measurements. These inaccurate measurements can impact the assessment of recently completed paving projects, or they can impact network-level measurements during routine assessment cycles. While these inaccuracies have differing impacts, they concern the agency, the data collector, and the paving contractor. Advancements in technology have created zero-speed or low-speed profilers, used to collect accurate data under stop-and-go conditions. Within this study, researchers collected profile data with a slow-speed profiler shadowing a conventional profiler at over 30 field locations with stop-and-go conditions. Various testing regimens were used that allowed researchers to compare the performance of each type of profiler, the repeatability of each type of profiler, and the lane influence on each type of profiler. The study indicates that slow-speed profilers outperform conventional profilers for accuracy and repeatability under stop-and-go conditions.

Abstract:
This presentation will discuss a multi-faceted and holistic approach to friction management. It will include the various roles/functions within an agency needed to collaborate with each other for success as well as a lifecycle view of friction on our paved surfaces. It will highlight developing technologies to predict friction response from non-contact measures, use of on-board vehicle systems to augment traditional response-based measures and explore current and upcoming research to evaluate functional performance of surface mixes in the laboratory as well as optimal friction treatment strategies.
Attendees will leave with a foundational understanding of the mechanisms between the highway surface and pneumatic rubber tires that affect both dry and wet friction. Similarly, they will have a better understanding of what friction management is and how it relates to highway safety with the goal of reducing crashes, particularly serious injury and fatality crashes. They will also learn what data may be readily available to help them recognize and respond to suspect friction locations. All culminating in a vision of what a mature highway friction management program can be in the near future.

Abstract:
The transportation pooled fund TPF-5(463): “Managing the Pavement Properties for Improved Safety”, has since its inception in 2006, held equipment comparisons to compare of the results of different surface properties measurement equipment. Last year, the pooled fund held a Macrotexture Network Level Measurement Comparisons at the Illinois Certification and Research Track (ICART). The objective of this equipment comparison (Rodeo) was to evaluate the performance of several different line laser devices that are used for network level macrotexture data collection.
This presentation will explain the characteristics of all the equipment that participated in the Rodeo and the methodology used to make the comparison. The equipment included 2 walk-behind and 4 high-speed devices that will be described in more detail in a Rodeo Report that will be placed on the website. The methodology used to do the comparison is “Limits of Agreement”. A summary of the results will show data for each of the three lanes in ICART (CRCP, asphalt, and JCRP, respectively). Overall, the results of the comparison were very satisfactory having learned from this initial experience, hoping to redo it in 2026 having fixed some of the faults identified in this first round of testing. It is the intention of the pooled fund that this process will lead to complete qualifications (repeatability and reproducibility) that will allow a future certification of the devices used in the comparisons made by the TPF-5(463) Pooled Fund.

Abstract:
A historical perspective of pavement condition assessment and its use from a 40 year highway engineer’s viewpoint. How and who are the key factors that move a concept from “”state of the art”” to “”state of the practice”” to eventually “”standard practice””. Many times it comes down to a few individuals that have the insight and perseverance. It always includes information/data quality management. RPUG has been the central hub for many pavement condition assessment standard practices and the nations roads are better for that.

It is hopeful that the perspective shared will be of interest to the majority of RPUG attendees and in particular the newer ones into our industry.

Abstract:

Abstract:
Network-level pavement condition surveys using 3D imaging systems are typically executed using fixed configuration settings selected to provide good results across diverse pavement types. For project-level analyses, the parameters that determine the cracking detection can be optimized, but this approach is impractical at network scale due to time and cost constraints. As a result, default settings are routinely applied, despite known limitations on certain pavement types, which include jointed concrete, open-graded asphalt and chip seals.

This paper presents a production-oriented hybrid framework that integrates LCMS-based 3D distress detection with artificial intelligence to adapt processing configurations based on pavement type, while preserving operational reliability. Rather than replacing existing workflows, the proposed approach augments standard LCMS processing with an AI-driven decision layer that routes the processing library to an alternative configuration setting when warranted.

First, a framework was developed to analyze pavement imagery using multiple LCMS settings, enabling direct comparison and manual selection of optimal results for specific pavement types. Second, a convolutional neural network based on ResNet50 was trained using manually reviewed project data (LCMS imagery) to identify pavement surfaces that need to be processed with different configuration settings.

To support deployment in production, three operational modes are implemented: an exhaustive-processing approach where every section is processed using all groups of configuration settings; an assisted approach in which only the AI-flagged sections are processed with both default and optimized settings; and an expert approach where AI-selected settings are applied exclusively.

Final quality control remains mandatory, with all corrections logged to continuously expand the training dataset and introduce new configuration groups. This feedback-driven framework enables measurable improvements in distress detection on challenging pavement types while maintaining production efficiency and data quality.

Abstract:
Three-dimensional pavement surface measurement systems are increasingly used to provide network-scale condition data for modern asset decision-making. However, the value of these systems depends not only on sensor capability but also on certified data-collection standards, transparent processing rules, and defensible verification methods that demonstrate that outputs reflect true in-service distress. This abstract presents a practical framework for three-dimensional surface measurement technology, certification, and automated pavement distress identification and reporting, supported by a verification case study for Toowoomba Regional Council using Laser Crack Measurement System (LCMS) data.
The framework integrates three pillars. Technology: high-density three-dimensional surface capture with right-of-way imagery enables objective measurement of rutting and surface-texture loss at the wheel-path scale, beyond subjective visual ratings. Certification: an auditable assurance layer defining minimum requirements for sensor calibration, survey coverage, lane referencing, processing configuration control, and quality checks. Evidence includes calibration records, repeatability checks, processing version control, and independent verification using imagery supported by targeted field review. Automated distress identification and reporting: consistent algorithms convert the three-dimensional surface into repeatable distress metrics for reporting, trend analysis, and treatment prioritisation.
The case study examined a targeted verification exercise after network-level LCMS routing values were observed to exceed those previously reported in an earlier LRMS survey. Segments with the largest variance were selected to assess whether the change represented data error, sudden pavement deterioration, or a methodological shift. Verification compared the rut-depth calculation methods and reviewed the right-of-way and laser pavement imagery. The review found LCMS processing applies lane boundary designations before rut depth calculation, enabling rutting to be computed within adjusted lane boundaries and defined wheel paths. In contrast, LRMS rutting was calculated across a 3.5 m lane width, irrespective of geometry or wheel path location, which can dilute results by averaging non-trafficked areas. Segment checks matched the LCMS values to the observed deformation.

Abstract:
Pavement Management is an important part of the work various DOTs have to do to keep their pavements in safe, smooth and operational conditions for the sake of the road users. Functionally, a lot of focus has been given to IRI and also is used a lot in treatment decisions. Data collection is done with profilometers that measure the deviation from smoothness based on the QC model but are usually expensive and collection is done annually. Recent technological advancements have made it possible for continuous large-scale road data collection from connected vehicles utilizing on-board vehicle sensors. We utilize the data from NIRA for ride quality to look at the road roughness values for Richmond (2000+ miles) for network-level analysis for the year 2023 and utilize temporal data from 2022-2024 for analyzing seasonal variations in growth and also compare CoV (%) for individual sections within a year with structural information obtained from TSD for around 200 miles.

Results from the network-level analysis, an R² fit of 0.67 and overall classification accuracy of 66% is observed. Based on different road classifications we observe an R² fit of 0.73, 0.56 and 0.66 for Interstates, Highways and Secondary roads along with classification accuracy of 72%, 61% and 73% respectively and can reach up to an R² of 0.87 for uniform road sections. Analyzing seasonality in trends from monthly data, it is observed that average road roughness grows strongly from winter to spring and peaks in summer and remains stable or declines slightly during fall. From the functional-structural analysis, the CoV (%) shows a stronger positive relationship with SCI300 (mils) sections for the high (90th Percentile) CoV (%) sections as compared to sections with low (10th Percentile) CoV (%) with the Beta (Impact Coefficient) being 5-6 times greater.

Abstract:
Road raveling, characterized by the progressive loss of aggregate from asphalt pavements, is a widespread surface distress that negatively affects safety, ride quality, and pavement life. Although most transportation agencies rate raveling severity on a scale from 0 to 3, there is substantial inconsistency in how severity levels are defined. For example, TxDOT relies on percentage-based criteria, while FDOT and GDOT primarily use qualitative textual descriptions. Despite advances in sensing technologies, agencies still depend on manual visual inspection for raveling assessment.

Recent adoption of 3D laser pavement measurement systems has enabled large-scale numeric data collection; however, current machine learning approaches typically output a single severity score with limited interpretability. These methods can also be confounded by coexisting distresses, such as cracking, which frequently appear alongside raveling. This lack of transparency and robustness limits trust and operational adoption.

This paper presents a Vision-Language Model (VLM)–based framework for automated raveling assessment that integrates visual understanding with natural-language reasoning. The proposed approach uses annotated pavement images derived from 3D laser data collected on GDOT roadways to predict raveling severity while simultaneously generating human-readable explanations that describe the visual characteristics influencing each prediction. The framework also demonstrates the potential for language-based recommendations of candidate maintenance or treatment strategies based on observed surface conditions. Multiple large-scale VLM architectures, including LLaMA-4 (128×17B) and Gemma-3 (27B), are utilized to evaluate the feasibility of this approach.

The proposed methodology aims to improve the transparency, robustness, and interpretability of automated raveling assessment, providing a pathway toward more consistent and explainable pavement distress evaluation across agencies.

Abstract:
In this presentation we introduce a new system for 3D surface measurements based on stereoscopy and compare it with existing well proven technologies. It consists of two line-scan cameras and a light system that records two stream of pictures of the surface at slightly different viewing angles. With a matching algorithm the entire surface can be reconstructed to generate a 3D surface scan. The system is intended for various profiling measurements such as texture and lane-marking thickness, or IRI, Cross-slope and rutting when combined with inertial sensors.
Som advantages of vision based surface scans include high line rate compared to line lasers, and high quality surface images which can be used for ex. crack detection.
Macro texture measurements of roads are compared with traditional point based profiling lasers and line lasers. The results shows both areas of high agreement between the sensors but also areas where different texture readings are observed. The areas that has higher difference might be due to measurement inaccuracy or that spectral responses of the different sensors are not exactly the same. Repeatability measurements are also performed showing with very high agreement between texture measured at two independent runs over the same road. Full surface scans are also presented showing the rich level of details that can be recorded by this technology.

Abstract:
Advances in Every Speed Profiler (ESP) technology enable pavement roughness data collection on roadway segments historically excluded from network-level International Roughness Index (IRI) reporting. These exclusions are primarily associated with limitations of traditional high-speed inertial profilers under low-speed operation, sudden speed changes, stop-and-go traffic, and other conditions that hinder continuous and reliable roughness measurement.

This study evaluates network-wide pavement roughness using a large-scale dataset collected simultaneously with an ESP and a High-Speed Inertial Profiler (HSIP) installed on the same survey vehicle. The dataset spans both a state DOT network and a major municipal roadway network.

The analysis first quantifies the spatial expansion of IRI coverage enabled by ESP by characterizing the extent, spatial distribution, and roadway characteristics of segments newly captured in locations where HSIP data collection is often limited or infeasible. A pairwise comparison of ESP- and HSIP-derived IRI values is then conducted on commonly surveyed roadway segments to evaluate inter-system agreement, bias, and variability under comparable operating conditions.

Next, roadway segments captured exclusively by ESP are analyzed to examine pavement roughness behavior and characteristics in portions of the roadway network that cannot be reliably surveyed using HSIP systems. Finally, a network-level comparison is performed to evaluate differences in reported IRI statistics and condition classifications when using ESP-only versus HSIP-only data.

Overall, the results demonstrate that ESP data provide more complete and representative pavement roughness information and that profiler selection can meaningfully influence network-level condition metrics and performance assessment. The findings offer practical insights for agencies considering expanded IRI reporting across complex roadway networks.

Abstract:
Since opening in 2023, the Illinois Certification and Research Track (ICART) has provided states and contractors a space to conduct valuable research, operate government programs, and test equipment in a controlled environment without the hazards of live traffic. One of IDOT’s more significant programs is the Profiler Certification Program. The program requires the participant to collect inertial profiler data over several test sections of varying pavement types and compares their results against a reference profiler. This ensures that the inertial profiler system is collecting accurate data in accordance with AASHTO R56 and gives confidence to contractors and state entities that quality data is being provided. Since the opening of ICART, we have seen significant growth in the number of certifications with 2025 having over 100 certification appointments. With growth brings new challenges along with opportunities to refine our processes and increase efficiency. This presentation will showcase where ICART and its certification program began, where they are today, and what we can expect in the future.

Abstract:
Austroads is the association of the Australian and New Zealand transport agencies, representing all levels of government. One of its functions is to develop and maintain technical guides and tools to promote a nationally consistent approach to the design, maintenance and operation of road networks. This includes overseeing the development of test methods for pavement condition data collection, including roughness measurement.

The original roughness test methods for data collection and validation test methods were first published in 2007 and since then have been revised on a regular basis. The revision process includes comparing the existing test methods with the latest international standards, including AASHTO, to take advantage of different requirements that are presently not covered by the Australian test methods, but whose inclusion would make the current test methods more robust. Another reason for regular revision is to ensure that the test methods cater for any new technological developments.

The Austroads test methods are presently being revised, and this presentation will focus on the key similarities and differences between the Austroads and AASHTO documents discussing the pros and cons of each plus some of the recent changes due to technological developments in the field of roughness measurement. The presentation will also provide an update on the status of the Australian roughness certification process for laser profilers.

Abstract:

Abstract:
ProVAL has been the standard software platform for longitudinal profile and ride quality analysis for agencies and industry since 2001. This presentation will describe the brief history of ProVAL development, current key functions, and the next round of enhancements in 2026 and 2027. The new enhancements will include: 1) Support Speed Profiles and Stop-and-Go Exclusions, 2) Support the Analysis for AASHTO R56-25 profiler certification, 3) Support for Stationing, and 4) Miscellaneous Features for Improvements. The presenter will also solicit the audience’s feedback on the GUI design and functionality of the new enhancements and ensure they meet users’ needs.

Abstract:

Abstract:
A mathematical description of a road surface should have several characteristics; it should be easily understood, compact when digitally stored, unambiguous, and, importantly, it should represent multiple attributes (height, color, friction) at multiple scales. The objective of this work is to develop a road description that streamlines both TPP data acquisition and certification analysis.

For each road a local Reference Coordinate System is established, the origin of which is defined in terms of global latitude and longitude. This Reference Origin is also the starting point for the Reference Line. Points along the Reference Line are defined by their local Easting and Northing coordinates. The Reference Line serves as the local datum for horizontal locations along the road and, as such, it is typically the Lane Center or some nominal road centerline. Distances along the Reference Line as measured from the Reference Origin are identified by U, the longitudinal coordinate of the Reference Line. The transverse coordinate that remains perpendicular to the Reference Line, regardless of the curvature of the Reference Line, is V.

Attributes such as elevation, color (RGB), and friction are then defined on a particular Grid and each Grid is defined relative to the unique Reference Line for the road. The Reference Coordinates U and V are continuous, but Attributes are recorded at discrete locations: Nodes. Grids are the means by which this discretization is defined and since different discretizations may be necessary for different Attributes, there may be a different Grid for each Attribute (e.g., elevation at one-centimeter, friction at one-meter intervals). Of particular interest are 2D Grids that can be alternately considered a longitudinal sequence of transverse profiles or a transverse sequence of longitudinal profiles. The advantages of this road description is demonstrated through examples of cross-slope, crowning, and rutting.

Abstract:
Accurate estimation of roadway cross-slope and longitudinal grade is critical for pavement performance, drainage, and safety analyses. This study presents a systematic literature review of methods used to calculate cross-slope and grade from diverse data sources, including traditional surveying techniques, inertial profilers, photogrammetry, mobile mapping systems, and airborne LiDAR. The review synthesizes how different studies define analysis units, extract roadway surfaces, handle noise and outliers, and estimate slope using profile-based, surface-fitting, or regression approaches. Strengths and limitations of each data source are discussed in terms of accuracy, spatial resolution, scalability, and applicability to network-level versus site-specific analyses. Key gaps are identified related to consistency in methodology, validation practices, and treatment of complex roadway features such as superelevation transitions. The findings provide a structured foundation for selecting appropriate slope estimation techniques in modern transportation infrastructure analysis.

Abstract:
The presentation describes an algorithm for detecting curl and warp using longitudinal profile measurements. The algorithm applies a combination of filters that retain periodic content associated with the slab length and integer fractions of the slab length (i.e., the slab-related roughness). Subtraction of the filtered profile from the unfiltered profile isolates the content that is not associated with slab effects (i.e., background roughness). The algorithm is suited for roughness analysis on network-level profiles and long project-level profiles, because it runs efficiently and does not require identification of joint locations or any slab-by-slab analysis. However, algorithm is not suited for analysis on profiles of short segments or identification of isolated curl and warp, because it average features over many slabs.
The use of the algorithm to track changes in slab-related roughness caused by environmental changes is demonstrated. The presentation also demonstrates the use of the algorithm to study the relative effectiveness of diamond grinding on reducing slab-related roughness and background roughness using profile measurements before and after grinding under various environmental conditions.

Abstract:

Abstract:
Hydroplaning poses a significant threat to traffic safety, occurring when water depth on a roadway exceeds the pavement’s drainage capacity, causing a loss of tire-pavement friction. Traditional risk assessment methods often rely on deterministic models with fixed input values, failing to account for the inherent variability in rainfall intensity, pavement geometry, and tire characteristics. This research proposes a probabilistic framework to assess hydroplaning potential on Texas highways using Monte Carlo Simulation (MCS). The study focuses on dynamic hydroplaning in passenger cars, utilizing Gallaway’s empirical equation to model water film thickness and predict hydroplaning speeds. To ground the analysis in real-world safety data, study segments are identified by filtering police crash narratives for keywords indicative of hydroplaning, such as “skidding,” “loss of control,” “standing water,” and “hydroplaning.” This approach ensures the simulation focuses on sites with a documented history of hydroplaning-related failures, establishing a robust foundation for the subsequent probabilistic analysis.

To model these locations with high precision, the framework integrates high-resolution Light Detection and Ranging (LiDAR) point cloud data to extract in-situ roadway geometry, specifically cross slopes and longitudinal grades. These inputs are coupled with pavement macrotexture and the Bartlett-Lewis Rectangular Pulse Model to generate realistic, short-duration rainfall intensities. By simulating thousands of scenarios, the study calculates the probability of hydroplaning speeds falling below posted limits. By quantifying risk as a probability rather than a fixed value, this research provides transportation agencies with a nuanced tool for identifying hazardous hotspots, offering a more realistic approach to infrastructure management and the reduction of wet-weather accidents.

Abstract:
This presentation will put forth the conceptual design for a vehicle and trailer system to evaluate the contribution of the pavement surface to skid resistance. It will not only outline the major functional systems but also explain the rationale behind the choices that are made. As in the past, its overall performance will still be constrained by the complexity of the tire-pavement interaction. For example, factors such as test speed, temperature, test tire tread depth will need to be measured and controlled. The potential improvement provided by the advances in technology over the past 70 years, both electromechanical and electronic allows for a more complete and accurate measure of this complex and important metric. The goal of the proposed system is to provide a more complete and accurate data set for evaluating the pavement relative to its contribution to safe and predictable vehicle control.
This is an intermediate step that is needed now while the research is accomplished enabling the measurement of the surface features on a network basis and translation of them into skid resistance performance. This device will be important for collecting a baseline for evaluating the values derived from the surface micro and macro profiles.

Abstract:
This study extends prior development of the estimated available pavement friction at operating speed, FRS(v), by quantifying its crash elasticity and facility-specific sensitivity within Safety Performance Function (SPF) models. The FRS(v) metric integrates low slip-speed friction and pavement macrotexture to represent available tire–pavement friction under realistic operating conditions. Speed-dependent friction values at 40, 55, and 70 mph were estimated using the ASTM E1960-07 speed-correction procedure and incorporated into negative binomial SPFs for freeways, rural two-lane two-way highways, and urban and suburban arterials. Beyond conventional model evaluation based on goodness-of-fit metrics, this analysis quantifies the percentage change in expected crashes associated with one-standard-deviation increases in FRS(v), enabling scale-independent comparison across operating speeds and roadway functional classes. Results indicate that both the magnitude and stability of the friction–crash relationship vary substantially by facility type and operating speed. On freeways, a one-standard-deviation increase in FRS113 (70 mph) is associated with approximately a 6–7% reduction in crashes, representing the strongest and most stable elasticity among freeway models. For rural two-lane two-way highways, the largest elasticity is observed for FRS65 (40 mph), corresponding to an estimated 9% reduction in crashes, with diminishing effects at higher speeds. Urban and suburban arterials exhibit substantially larger estimated elasticities, up to approximately 20% per standard deviation, but with reduced model stability, reflecting heterogeneous speed distributions and geometric characteristics. A supplementary network-level application demonstrates how elasticity-based FRS metrics can support investigatory friction thresholds, segment prioritization, and safety-oriented pavement management decisions. The findings show that elasticity-based interpretation of speed-adjusted pavement friction provides a more informative and transferable safety metric than friction or macrotexture evaluated independently, supporting proactive and data-driven pavement friction management.

Abstract:
Skid resistance is critical for roadway safety but is typically measured using contact-based methods with reliability and operational challenges. Pavement texture offers a complementary skid characterization, yet current high-speed sensing is mostly limited to macro-texture while micro-texture remains confined to static/lab settings. This study introduces a framework for predicting pavement safety performance by incorporating micro-texture information generated through super‑resolution reconstruction. Paired macro‑ and micro‑texture data were collected from ten roadway segments at low and high speeds using a mobile platform equipped with a 0.1‑mm‑resolution 3D safety sensor and a 1‑mm‑resolution highway‑speed profiling system. Four generative models—GAN, Diffusion, Flow Matching, and Fractal‑based—were trained to reconstruct micro-texture from macro-texture inputs, and their performance was evaluated using mean squared error (MSE), peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). The best‑performing model was then used to compute texture indicators derived from height‑based, spatial‑based, feature‑based, and hybrid approaches. Super‑resolution texture data were also generated for 26 additional highway segments in Oklahoma and used to compute their corresponding texture indicators. These indicators, combined with historical crash data, supported the development of texture‑enhanced Safety Performance Functions (SPFs) using Poisson, Negative Binomial, Zero‑Inflated Poisson, and Zero‑Inflated Negative Binomial models. The proposed framework enables high‑fidelity, texture‑informed safety assessment at highway speeds and offers a practical, data‑driven approach for pavement safety management.

Abstract:
The collection of continuous tire-pavement friction data and its use in pavement materials, maintenance, and safety decision-making is a relatively recent development in the United States. As collection and implementation of continuous tire-pavement friction data become increasingly prevalent, a critical next step is developing a framework that includes substantive data quality standards and guidelines for data management practices to monitor and evaluate data production and give agencies the confidence to make informed decisions.
To support the progression of high-quality data from data collection to use, this presentation will borrow a data value chain framework that treats data as a raw material of decision-making where value can be added – or diminished – at each stage of production (“The Data Value Chain: Moving from Production to Impact,” Open Data Watch, 2019). At each of the five (5) proposed stages – Data Collection, Validation, Integration, Analysis, and Use – the authors will identify key processes, activities, and success criteria that together result in a repeatable process with built-in feedback mechanisms to ensures both high-quality inputs and high-quality products.
Using a specific use case to illustrate a potential friction data value chain, the authors will detail concrete activities such as equipment validation, specify potential data quality gates with measurable criteria, and identify the use and benefits of third-party data in the process of creating value with continuous tire-pavement friction data.

Abstract:

Abstract:
Transportation agencies rely on concrete pavement joint condition and performance data to support maintenance planning, structural evaluation, and investment decisions. However, accurately accounting for longitudinal and transverse joints remains a challenge, particularly as Traffic Speed Deflection Devices (TSDDs) are increasingly used for network-level pavement assessment. Concrete joints can introduce localized anomalies in deflection measurements, reducing the signal-to-noise ratio and increasing uncertainty in pavement data interpretation when joint locations are not explicitly identified or accounted for.

As TSDDs gain adoption across highway and airport pavements, agencies face a growing need for scalable, objective methods to identify and spatially reference concrete joints and integrate that information directly with deflection datasets. Manual interpretation and rule-based approaches are difficult to scale, inconsistent across networks, and costly to maintain.

This presentation introduces a novel machine-learning (ML) based solution that automatically detects and groups longitudinal and transverse concrete joints using ultra-high-resolution 3D pavement data. By accurately mapping joint locations and enabling their overlay with deflection measurements, the approach provides pavement engineers with improved context for data interpretation and helps reduce uncertainty in structural analysis.

Moving beyond AI hype, the session presents a practical, real-world application of modern convolutional neural networks (CNNs) developed specifically for pavement engineering use cases. The presentation highlights key design trade-offs, implementation challenges, and performance considerations, including the role of GPU-accelerated processing. Presented from the perspective of a former data scientist and AI developer, the session emphasizes conceptual understanding over mathematical detail, helping transportation agency practitioners better evaluate where Vision AI can deliver measurable value within pavement management workflows.

Abstract:
The Pavement Condition Index (PCI) is a cornerstone of pavement management systems, yet traditional PCI surveys rely on manual visual inspections that are time-consuming, costly, and subject to human errors. These limitations make frequent, network-level condition assessment difficult. This study introduces a practical, vehicle-based framework for automating pavement distress detection and PCI evaluation using fused two-dimensional (2D) and three-dimensional (3D) image data. While automated distress detection has been widely studied, limited research has focused on integrated, field-ready frameworks that directly translate automated detections into PCI values consistent with ASTM D6433; this study addresses that gap by presenting a practical end-to-end PCI automation framework.

An instrumented vehicle collects synchronized 2D and 3D pavement images, which are fused and analyzed using a YOLO-based deep learning model to detect asphalt surface distresses, including transverse cracking, longitudinal cracking, alligator cracking, block cracking, and patching or utility cuts. Distress instances are manually annotated to develop the training dataset. Detected distresses are quantified at the pavement sample-unit level and translated into PCI values consistent with ASTM D6433, with severity assigned using crack width for linear distresses (longitudinal and transverse cracking) and distress density for non-linear distresses (alligator cracking, block cracking, and patching or utility cuts). Standard deduct-value procedures are then applied to compute PCI. An additional segmentation step within detected crack regions supports crack-width estimation for linear cracking.

The approach is demonstrated using asphalt parking lots, which provide a controlled yet realistic environment with diverse distress conditions. Automated PCI estimates are compared with limited manual reference PCI documentation to evaluate consistency and reliability. The results highlight the potential of fused 2D/3D imaging and deep learning to support faster, safer, and more objective pavement condition assessments, offering a pathway toward scalable and field-ready PCI automation.

Abstract:
Contemporary roadway management practice increasingly depends on multimodal, spatially coherent datasets while contending with constraints on mobilization effort, roadway access, and the availability of dedicated survey platforms. This study describes a vehicle‑agnostic data collection architecture that integrates pavement profiling, panoramic roadway imaging, and 360° oblique LiDAR within a unified, traffic‑speed acquisition workflow.
The system utilizes modular sensor pods deployable on standard fleet vehicles, enabling rapid mobilization and reducing operational constraints commonly observed in large‑scale network surveys. A unified timing and synchronization framework merges sub‑millimeter 3D pavement measurements with high‑resolution imagery and dense point‑cloud data, producing a spatially consistent digital representation of roadway corridors.
To maintain positional integrity in variable navigation environments, survey‑grade inertial navigation is combined with DMI‑aided correction, supporting accurate georeferencing through GNSS‑limited segments. This approach enables consistent derivation of pavement condition metrics and supports continuity in surface profiling across heterogeneous roadway and environmental conditions.
The collection architecture is paired with a cloud‑based processing pipeline designed to support retrospective analysis as computational methods evolve. This includes potential applications in automated surface‑condition assessment, asset extraction, and multimodal roadway condition inference using emerging AI‑based techniques.
Taken together, the system architecture demonstrates an approach for scalable, multimodal roadway data collection that reduces mobilization requirements while supporting high‑fidelity characterization of pavement networks and roadway corridors.

Abstract:
This research aims to develop a two-stage framework for the automated detection and quantification of cracks on asphalt and concrete pavements using pavement surface imagery. To overcome the limitations of traditional edge-extraction methods, the proposed approach enhances the identification of fine and thin cracks. In the first stage, the cracking is located within 2D/3D vehicle-captured images using bounding boxes generated by either human experts or predictive models such as a vision-language model developed in our prior study. The second stage utilizes high-resolution segmentation AI models to delineate crack boundaries and extract geometric features (width, length, and area), facilitating severity scoring according to distress identification guidelines and pavement management standards specified by different highway transportation agencies. We evaluate several state-of-the-art architectures, including YOLOv26, SCSegamba (a structure-aware vision mamba network), and Dichotomous Image Segmentation (DIS), by training and testing them on a new dataset of 1,432 pavement images. A morphological method is then applied to quantify the lengths and widths of cracks. Results indicate that this cascaded framework significantly improves detection accuracy and shape capturing for subtle surface cracks, providing a more quantifiable method for automated pavement condition assessment.

Abstract:
Accurate pavement distress segmentation with limited labeled data remains a major challenge for deep learning-based infrastructure inspection. This study proposes DAPDSnet (Difficulty-Aware Pavement Distress Segmentation Network), a semantic segmentation framework that incorporates pixel-level difficulty estimation to guide an active learning (AL) strategy for efficient sample selection. By characterizing spatial uncertainty at the pixel scale and aggregating it for sample prioritization, the framework iteratively selects informative unlabeled samples for annotation. Experiments on a publicly available Pavement-2K dataset developed by our team, consisting of over 3,000 3D pavement images, demonstrate that DAPDSnet achieves over 85% of full-supervision performance in terms of Mean Intersection over Union (MIoU) and Mean F1 score (MF1) using only 45% of the labeled data. The framework consistently outperforms random sampling, particularly in early training stages. Ablation studies validate the network design and identify MobileNetV2 as an effective backbone for AL scenarios. Visualization of the AL process illustrates how model uncertainty shifts toward structurally complex pavement distresses. The proposed framework reduces annotation requirements while preserving segmentation performance.

Abstract:
Pavement grooving & surface texture are critical to achieving safe aircraft operations in wet weather. Friction is often a key factor in aircraft accidents during landing and take-off. A laboratory testbed was designed and built to investigate the frictional interaction between aircraft tire and pavement surface. The testbed is built on a platform with a rotational arm to support an aircraft tire to travel and brake on a circular, reconfigurable track. Mechatronics design, data acquisition, and control system modeling of subsystems in the platform were presented and discussed. By designing the braking torque and the rotating arm motion, the testbed possesses the capability to emulate the dynamic characteristics of aircraft tire-runway interactions and braking maneuvers.

Experiments were conducted using the laboratory testbed to measure the friction coefficient of tire at various slip ratios across different grooved surfaces and water conditions. Finite element models are developed to simulate tire-water-pavement interaction to predict the friction coefficient between aircraft tires and wet grooved pavement. The aircraft tire model was validated by the contact footprint and load-deflection curve. The calculated friction coefficients were found to match well with field measurements. Through the developed approach, the evaluation of hydroplaning speed and braking distance of the landing aircraft can be performed.

The findings from laboratory experiments and analytical modeling show that the square and trapezoidal grooves on runway pavements can provide similar skid resistance to aircraft tire at wet conditions in general. The simulate results suggest that the trapezoidal groove may provide greater friction at high speeds and water film depths.

Abstract:
This study evaluates the effectiveness of laboratory polishing devices—the three‑wheel polisher and the Micro‑Deval—in replicating pavement friction and texture deterioration observed under real traffic and establishes quantitative relationships that translate laboratory cycles to field conditions. Although both devices are known to polish asphalt slabs and aggregates effectively, limited data exist linking laboratory polishing intensity to actual traffic loads. To address this gap, researchers paired field monitoring of asphalt and seal coat sections with laboratory testing on slabs fabricated from the same mixture designs. Field sites were repeatedly tested over the study period to document friction and texture evolution, while laboratory slabs underwent controlled polishing using the three‑wheel polisher, and aggregates were evaluated before and after Micro‑Deval conditioning. Analysis showed a strong correlation (R² = 0.74) between field friction measurements and predictions from the three‑wheel polisher, enabling derivation of a practical conversion factor of 312.5 cumulative vehicle passes per laboratory polishing cycle. Both field and laboratory samples exhibited decreasing friction with polishing, though low‑volume road sections (AADT < 1,000) showed minimal degradation even after extended service. Texture (MPD) trends varied by surface type, with seal coats showing MPD reductions and most other surfaces showing increases; laboratory texture predictions to field measurement were generally weaker due to construction variability and surface deterioration patterns. Aggregate ring tests using dynamic friction values provided stronger correlations to field (R² = 0.70) and laboratory (R² = 0.56) friction than traditional SAC A specifications. These findings provide TxDOT with validated methods and conversion tools to predict long‑term pavement friction performance and support improved aggregate selection and mix design decisions for wet‑weather safety.

Abstract:
Proactive roadway safety management increasingly recognizes pavement surface characteristics as critical contributors to reducing crash risk, particularly under high-speed and wet weather conditions. Within the Safe System framework, addressing friction demand requires consideration not only during maintenance treatments but also at the pavement design stage. Pavement macrotexture plays a fundamental role in sustaining tire–pavement contact, facilitating water drainage, and preserving friction at operating speeds where microtexture alone is insufficient. Despite its demonstrated safety relevance, macrotexture is rarely treated as a design-level performance target in conventional asphalt mixture design procedures. This study presents a data-driven framework for predicting pavement macrotexture, expressed as mean profile depth (MPD), using both statistical and machine-learning approaches. A comprehensive dataset of field and laboratory macrotexture measurements from recently constructed asphalt surfaces was compiled from multiple state transportation agencies. This research contributes to an ongoing national initiative led by the Federal Highway Administration (FHWA) to develop calibrated models capable of predicting as-constructed macrotexture from asphalt mixture design and aggregate properties. Predictor variables included aggregate gradation characteristics, mixture volumetrics, and surface mixture classifications. Linear regression models were first developed to establish interpretable relationships between mixture properties and MPD. These models were complemented by Random Forest and Bayesian-optimized eXtreme Gradient Boosting (XGBoost) algorithms to capture nonlinear interactions and complex dependencies. Results show that while linear models provided transparent engineering interpretation, machine-learning models improved predictive robustness and reduced error by more effectively handling multicollinearity and nonlinear effects. The combined modeling strategy demonstrates that pavement macrotexture can be reasonably predicted from asphalt mixture design parameters. This capability supports the consideration of macrotexture as a performance indicator during mixture design and construction, enabling agencies to make more informed decisions regarding material selection and mixture optimization, and to better integrate macrotexture targets within broader pavement friction management strategies to enhance roadway safety.

Abstract:
Accurate estimation of asphalt surface layer modulus is essential for pavement structural evaluation and network-level pavement management. Falling Weight Deflectometer (FWD) testing provides a non-destructive means of characterizing pavement response; however, conventional modulus backcalculation procedures are often sensitive to assumptions related to material properties, temperature, and pavement structural configuration. This study proposes a data-driven framework for predicting asphalt surface layer modulus using a comprehensive FWD dataset from Texas Long-Term Pavement Performance (LTPP) sections. The proposed dataset integrates peak deflection basin measurements, surface and base temperature data, applied drop load, and asphalt layer thickness to capture both structural and environmental influences on asphalt surface layer modulus. Multiple predictive modeling approaches will be developed to examine the ability of linear and nonlinear methods to represent relationships between FWD measured pavement response parameters and backcalculated surface layer modulus. The modeling approaches will include Multiple Linear Regression (MLR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and an Artificial Neural Network (ANN). Model development and evaluation will be performed using independent training and testing datasets to ensure robust comparison across modeling methods. Model outputs are intended to provide insight into the relative contribution of deflection basin features, temperature variables, and asphalt layer thickness in data-driven surface layer modulus prediction. This study would help to evaluate the practical applicability of data-driven approaches for interpreting FWD measurements and supporting pavement structural assessment. The proposed framework is expected to support more reliable use of FWD data for estimating asphalt surface layer modulus and to offer a practical tool for network-level pavement structural evaluation, contributing to improved pavement management and decision-making.

Abstract:
Effective road asset management relies on the accurate characterization of pavement distress. However, a major challenge remains in converting high-speed automated detection data into actionable maintenance engineering decisions. This study presents a hierarchical framework designed to bridge the gap between raw survey imagery and refinement maintenance categorization. To manage the massive data volume of national-scale surveys—which can exceed 350 million images annually—the framework utilizes a Deep Grid Segmentation model, known as αCrack, as its high-throughput detection backbone. This model ensures rapid identification of cracking and repair zones while optimizing the balance between detection precision and manual labeling costs. Building on this robust detection stage, the framework introduces an instantiation algorithm that refines basic binary results into detailed engineering categories. A key component of this process is an attention-based fusion module that integrates semantic features with geometric descriptors. This enables the system to distinguish between complex distress patterns, such as transverse, longitudinal, alligator, and block cracking. To further improve training stability and classification accuracy, a pseudo-label generator is integrated to provide reliable supervision signals during the model’s refinement phase. Experimental results show that this multi-stage method can be trained using coarse-grained classification labels to achieve fine-grained multi-class identification. By quantifying the density and specific morphology of these particular diseases, the framework successfully meets the engineering application requirements of achieving high-precision recognition with low labeling costs. This work shifts the technical focus from simple disease identification to more comprehensive diagnostic analysis, providing a scalable and accurate solution for intelligent road management that can meet the stringent requirements of modern production environments.

Abstract:
To adapt to the trend of road inspection shifting from macroscopic census to refined and in-depth detection, the industry’s demand for multi-dimensional and high-efficiency data has been increasing. However, existing inspection equipment generally have the drawbacks of a single detection dimension, low efficiency, and inconsistent spatiotemporal baselines of multi-source data, making it difficult to collaborate. Therefore, this research has developed a highly integrated on-board multi-sensor integrated rapid detection equipment to enhance detection efficiency and solve the problem of data fusion. The equipment overcomes two key technologies: one is the deep integration of multiple sensors, and the other is the synchronous data fusion. Specifically, it integrates four functional modules: the three-dimensional pavement detection module, which can collect pavement texture and surface defects with millimeter-level accuracy; the three-dimensional radar and vision integrated module, which simultaneously acquires the three-dimensional coordinates and visual information of roadside facilities such as guardrails and signs; the three-dimensional ground-penetrating radar module, which is used for non-destructive detection of hidden diseases such as cavities, voids, and water accumulation inside the road; and the retroreflective detection module, which precisely measures the optical performance of road markings. To address the problem of multi-source heterogeneous data fusion under high-speed movement, the equipment adopts high-precision synchronous timing technology based on hardware timestamps, combined with unique mechanical design and control algorithms, to ensure that all data have strictly unified spatiotemporal baselines, achieving “one-time passage, comprehensive perception”. After systematic joint debugging on actual roads and multi-scenario tests, this equipment demonstrates excellent engineering stability and data reliability. This achievement successfully develops a set of technologically advanced and functionally complete on-board integrated rapid detection equipment, achieving efficient, synchronous, and high-precision integrated collection of road surface, roadside facilities, and internal structural conditions, providing an advanced data collection method that can be promoted for engineering application for the digital and intelligent management of road assets, and has significant practical application value.

Abstract:
Approximately 40,000 roadway fatalities occur annually in the United States, with nearly 20% associated with insufficient tire-pavement friction. Pavement surface texture is a primary contributor to friction. Highway agencies usually rely on high-speed laser scanners to measure texture at the network-level. While this equipment could efficiently survey road networks, this technology is costly, and pavement lane coverage may be limited. Image-based 3D reconstruction (photogrammetry), which recovers 3D data from a series overlapping images has been used for accurate and cost-effective surveys of pavement surface characteristics. Although many studies have concluded that photogrammetry is a viable alternative to laser-based equipment for network-level monitoring of pavement texture, its use has been limited to stationary surveys on small-scale patches. In addition, no device exists for high-speed data collection. In this work, a prototype device for network-level texture measurement using photogrammetry is proposed. To ensure high-quality data, two industrial cameras shooting 25 Megapixel images up to 100 frames per second were utilized. The scale ambiguity was resolved without the need for ground control points by georeferencing images with high-accuracy location data from an RTK GNSS receiver. Furthermore, the cameras and GNSS receiver were electrically synchronized to minimize matching and georeferencing errors. The equipment was mounted on a van, and data were collected at a speed up to 40 mi/hr on local roads and at the Illinois Certification and Research Track (ICART). Texture data, expressed as Mean Profile Depth (MPD), collected by the proposed prototype and an AMES 9400HD laser scanner were in agreement and highly correlated.

Abstract:
Lane-departure crashes are a major safety concern on highway facilities, particularly under wet weather conditions when surface water reduces available skid resistance. While traffic volume and geometry are traditional safety predictors, the specific contributions of surface texture and friction, especially their combined interaction, have not been as closely evaluated. This study developed facility-specific safety performance functions (SPFs) to evaluate the influence of macrotexture and friction on wet-lane departure crashes across North Carolina’s primary road network.

Network-level data were collected by the Sideway-Force Coefficient Routine Investigation Machine (SCRIM). SPFs were developed for pavements grouped by operational speed (high speed, ≥ 60 mph, or low speed, < 60 mph) and facility type (divided or undivided). In addition to these groupings, roadway sections were further categorized according to the presence (or not) of interchanges. The models predicted wet lane departure crashes over a 13-month period by integrating mean profile depth (MPD), SCRIM Coefficient (SC), traffic exposure (AADT), and section length (L). This period included 6 months prior to the surface condition measurement, the month of measurement, and 6 months after the measurement. Although crashes are rare events and typically require 3–5 years for analysis, pavement friction and macrotexture are time-dependent and may vary on shorter time scales. An interaction term between macrotexture and friction (MPD × SC) was also included to capture the combined skid-resistance mechanism influencing wet-weather safety. Pavement macrotexture showed a statistically significant negative association with crash frequency for divided high-speed, interchange, and non-interchange facilities. The interaction between macrotexture and friction is statistically significant for divided high-speed and interchange facilities, demonstrating that the safety benefit of friction depends on the available macrotexture. Overall, findings support integrating friction and texture into network-level pavement management and safety analysis, particularly at high-speed and interchange facilities where surface-related crashes are most pronounced.

Abstract:
Numerous studies have shown strong correlations between laboratory friction measurements obtained using the Dynamic Friction Tester (DFT) and field friction measurements from the Locked Wheel Friction Tester (LWFT) with a ribbed tire. Individual correlation studies commonly report R2 results exceeding 70%, reinforcing the perception that laboratory friction testing can be used to predict field friction performance. However, substantial variability exists among the reported regression coefficients relating DFT and LWFT results, limiting the DFT’s usefulness in mix design to predict asphalt pavement friction.
This presentation synthesizes existing DFT–LWFT correlation studies and demonstrates that the observed variability is largely attributable to differences among device pairings rather than weak physical relationships. Both the DFT and LWFT exhibit inherent variability and are not perfectly reproducible across units, resulting in each unique DFT–LWFT pairing producing a distinct, yet internally consistent, linear relationship. To address this issue, a population-based modeling framework is proposed to characterize the family of DFT–LWFT relationships observed across studies. The framework jointly accounts for variability in regression slope and intercept, recognizes legitimate differences among studies and devices, and avoids disproportionate weighting of individual datasets based solely on sample size. By treating individual correlation studies as realizations from a broader population of possible device pairings, the approach preserves the underlying physical relationship between laboratory and field friction measurements while accommodating observed variability.
Preliminary results indicate that while a global relationship yields wide prediction intervals, device-specific correlations substantially reduce uncertainty. For a given DFT value, prediction intervals associated with a specific DFT–LWFT pairing are reduced by approximately 60% compared with a global model. These findings suggest that laboratory friction testing can be effectively incorporated into tire–pavement friction measurement and analysis when supported by device-specific calibration, providing agencies with a practical path to link laboratory measurements to field friction performance.

Abstract:
This study evaluates and optimizes roadway cross-slope estimation using both mobile LiDAR and static airborne LiDAR datasets. High-resolution point clouds from mobile mapping systems and publicly available airborne LiDAR (e.g., USGS QL2+) are processed to extract roadway surfaces and estimate cross-slope using patch-based plane fitting techniques. Results are compared across data sources to assess consistency, sensitivity to point density, and robustness to noise. The analysis demonstrates that mobile LiDAR provides finer spatial detail and improved accuracy at the lane level, while airborne LiDAR enables efficient network-scale assessments with acceptable precision. Optimization strategies, including patch size selection, orientation control, and noise filtering, are shown to significantly improve slope stability across both datasets. The study highlights trade-offs between accuracy and scalability and provides guidance for selecting LiDAR data sources for roadway safety and drainage applications.

Abstract:
Pavement surface texture is a multi-scale, spatially variable signal that strongly influences friction, yet many agencies still lack a repeatable, device-agnostic procedure to quantify spectral texture content, diagnose localized defects, and validate collection technologies against a defensible baseline while preserving links to friction. This paper presents SmartWave™—an analysis workflow that standardizes wavelet-based pavement surface texture analysis and technology comparison for pavement management and engineering applications. The study will illustrate the use of SmartWave™ to assess the characteristics and usability of different wavelet families used for discrete wavelet transform (DWT) and continuous wavelet transform (CWT) for 2D and 3D pavement surface texture data. The discussion will include the consistency of the parameters with observed friction measurements and the other established texture parameters.

For selected wavelet families (efficient wavelets), the study will discuss the impact of resolution, accuracy, precision, and noise of collected data on the estimated texture parameters including spectral and wavelets metrics. The evaluation is done using a sensitivity analysis performed in SmartWave™. Moreover, the study will present recommended settings for preprocessing steps including detrending, resampling, anti-alias filtering, gap/spike screening, and unit normalization. This evaluation will help in standardizing the preprocessing. Following the preprocessing the discussion will turn to evaluating the different data collection technology validity using paired baseline comparisons against a stationary reference device. The reference device provides a minimal variance and bias for texture measurements in the macro-texture and some portion of the micro-texture range.

The analysis in this study will be demonstrated using field datasets collected across multiple states to capture variability in materials, climate, and collection practices. Results show how SmartWave™ features strengthen interpretation of texture measurements and provide exportable, reproducible outputs that support harmonized texture measurement, transparent device qualification, and can lead to improved contactless friction evaluation.

Abstract:
Consistent and reliable crack reporting is essential for state highway agencies’ (SHAs) pavement management and Highway Performance Monitoring System (HPMS) submissions. However, existing crack definitions lack rigidity, leaving room for interpretation and leading to inconsistent reporting across agencies, sensors, and vendors. This paper presents a three-level cracking definition framework designed to eliminate such ambiguity through rigid definitions and precisely specified executions while preserving flexibility for diverse agency needs.

The framework is built upon the Crack Vector Model (CVM), which represents cracks as vectorized features composed of links, nodes, and vectors. All crack geometry, attributes, classification criteria, and aggregation rules are explicitly defined, leaving no room for subjective interpretation. This ensures that identical crack maps always produce identical reporting outputs. Based on this foundation, crack information is organized into three hierarchical reporting levels with rigid transitions between them:
• Level 1 captures fundamental link-wise crack information, including geometry, width, orientation, and location.
• Level 2 aggregates Level 1 data by crack type, severity, and zone for asphalt pavements or by slab for JPCP.
• Level 3 further aggregates into grid-based cracking percentages or slab states for network-level assessment. Because each higher level is derived strictly from the level below through explicitly defined rules, the entire reporting chain remains fully traceable and reproducible.

Experimental applications across asphalt pavement, CRCP, and JPCP demonstrate the framework’s ability to produce consistent results across different pavement types and agency practices. By combining rigid definitions and precisely specified executions with fundamental crack information, the framework offers SHAs a practical path toward consistent, repeatable, and comparable crack reporting, with the flexibility to support new reporting requirements as needed.

Abstract:
The Federal Highway Administration (FHWA) established a Use Stage Technical Subgroup (USTS) in 2023 to provide technical input on critical issues related to the pavement use stage. The focus of the USTS was on pavement-vehicle interaction (PVI) effects as they relate to pavement roughness, structural response, and texture, and how those in turn influence rolling resistance and vehicle fuel use or energy consumption over the pavement life cycle. The overarching goal of this effort was to provide highway agencies with additional information and tools to help guide and inform the pavement investment decision-making process.

The USTS included 15 members representing FHWA, State Departments of Transportation, academia, industry, and consultants, all with knowledge of PVI parameters and characteristics. These members participated in 10 meetings (between October 2023 and December 2024) and provided valuable input and feedback on critical factors associated with the pavement use stage. Specifically, the USTS provided the following outcomes: (a) a rationale for including PVI and how it affects vehicle energy consumption in a pavement life-cycle assessment (LCA) framework to support decision-making, (b) a list of the models/data/needs that are required for the most complete or most ideal inclusion of PVI in pavement LCA while taking into consideration the associated implementation requirements (i.e., feasibility and practicality considerations), (c) an outline for an approach (models/data/assumptions) to inform PVI decision-making today with available data, models, and tools, and (d) a compilation of pavement and other related key considerations, including data items and prediction models, that are still needed for the implementation of PVI considerations in the decision-making process.

This presentation summarizes some of the key takeaways and information that emanated from those meetings, along with a summary of current research needs and recommendations for short- and long-term implementation.

Abstract:
Crack detection and evaluation using 3D pavement surface imaging systems is widely adopted by state highway agencies. Although standards such as the AASHTO transverse pavement profiler (TPP) certification have been established to evaluate the overall quality of sensing data, there is currently no standardized approach for assessing the quality of crack-specific data. As a result, agencies often lack confidence in the reliability of cracking data used for pavement management.

This study develops a novel and practical methodology for evaluating cracking data quality by directly comparing test crack maps detected from sensing data with ground-reference crack maps obtained through manual field identification and labeling. Sensing systems are operated three times, both before and after manual crack labeling. White, washable paint is applied manually to augment crack visibility on selected pavement surfaces, typically over 20–50 ft, depending on cracking severity. This field procedure generally requires three raters and approximately two hours. The paint-augmented cracks are highly visible in pavement intensity images, facilitating ground-reference data annotation, and a drone-based RGB camera may be used to provide supplemental ground-reference imagery when available. The percentage of missing cracks is then quantified by comparing test crack maps with the ground reference.

Field tests and post-processing analyses were conducted, and the results indicate that up to 40% of cracking data may be missing and that alligator cracking may appear as longitudinal cracking when finer interconnected cracks go undetected, depending on the sensing system and pavement surface type. These findings demonstrate the feasibility of the proposed methodology as an effective and practical quality-control tool for assessing cracking data and supporting more reliable pavement management decisions.

Abstract:
Accurate and repeatable pavement profile measurement remains a cornerstone of pavement surface characterization, supporting ride quality evaluation, network-level pavement management, and agency certification programs. Conventional inertial profiler technologies, primarily based on point or limited multi-point laser sensors, provide longitudinal profile measurements but are constrained by sparse spatial sampling and sensitivity to surface texture variability.

This presentation explores solid-state area-scan LiDAR as an emerging pavement profile measurement technology and discusses its suitability for current and future applications under AASHTO R56. By capturing a continuous, high-density three-dimensional elevation surface across the pavement lane, area-scan LiDAR enables robust longitudinal profile extraction with improved repeatability and reduced sensitivity to localized texture, transverse features, and surface anomalies. The presentation outlines methods for deriving compliant profile measurements from 3-D elevation data while maintaining alignment with established inertial profiler certification criteria.

Beyond profile measurement, the same dense 3-D surface data inherently supports automated pavement distress identification, including cracking, rutting, faulting, and surface deformation. The presentation will discuss how profile-focused data collection systems using solid-state area-scan LiDAR can be incrementally upgraded through software-based processing and analytics to support distress detection, without fundamental changes to the sensing hardware or collection workflow. This approach enables agencies to preserve existing investments while expanding analytical capabilities over time.

By positioning solid-state area-scan LiDAR within established profile standards while highlighting its extensibility toward automated distress analysis, this presentation demonstrates how emerging 3-D measurement technologies can support both current certification requirements and future pavement surface characterization needs.

Abstract:
Accurate road alignment information is essential for autonomous driving, safety analysis, and digital infrastructure modeling. This study proposes a coarse-to-refined framework for extracting and parameterizing horizontal road alignment from unordered mobile laser scanning (MLS) point clouds. An adaptive Bézier vectorization method transforms dense, noisy curb points into smooth, continuous geometric representations. A hybrid identification strategy combining geometry-based segmentation and model fitting then detects standard design elements such as straight lines, arcs, and clothoids, and estimates their parameters. The method reduces threshold dependence and enhances robustness. Experiments on real-world data achieve over 99.4% correctness and completeness (0.5 m tolerance), while tests on simulated data yield 96.92 % mean Intersection over Union (IoU) and parameter errors below 0.401 %. Runtime evaluation confirms its efficiency for offline processing, averaging 245 s/km for geometry extraction and 21 s/km for alignment identification on a single-core CPU. The framework provides an accurate, scalable solution for infrastructure modeling using cost-effective MLS systems.

Abstract:
This is a TPF-5(463) Friction Pooled Fund Meeting to discuss the TPF’s friction studies.

All TPF members and friends are welcome!

Abstract:
This workshop is hands-on and focuses on ProVAL software hands-on exercises. It also includes a ProVAL users’ forum for interacting with users and gathering feedback. Register here (https://events.teams.microsoft.com/event/58c8d38f-da21-4a16-b9af-322ecb34b00e@8aec2bf0-04af-4841-bcf6-bac6a58dd4ef).

Trainers include George Chang & Dave Merritt (Transtec Group, A Terracon Company)

All participants need to bring their laptop computer preinstalled with ProVAL 4.0 and sample files. Download detailed instructions. (https://www.roadprofile.com/wp-content/uploads/2026/02/ProVAL-Workshop-RPUG2026-v1.pdf)

This workshop is open to all!