Accurate and efficient pavement condition assessment is essential for maintaining roadway safety and optimizing maintenance investments. However, conventional assessment methods such as manual visual inspections and specialized sensing equipment are often time-consuming, expensive, and difficult to scale across large networks. Recent advancements in generative artificial intelligence (GAI) have introduced new opportunities for automating visual interpretation tasks using street-level imagery. This study evaluates the performance of seven multimodal large language models (MLLMs) for road surface condition assessment, including three proprietary models (Gemini 2.5 Pro, OpenAI o1, and GPT-4o) and four open-source models (Gemma 3, Llama 3.2, LLaVA v1.6 Mistral, and LLaVA v1.6 Vicuna). The models were tested across four task categories relevant to pavement management: distress and feature identification, spatial pattern recognition, severity evaluation, and maintenance interval estimation. Model performance was assessed across five dimensions: response rate, response correctness, consistency, multimodal errors, and overall computational intensity and cost. Results indicate that MLLMs can interpret street-level imagery and generate task-relevant outputs in a cost-effective manner. Among the evaluated models, we recommend GPT-4o as the preferred option, as it balances responsiveness, accuracy, and computational cost.

Accurate and efficient pavement condition assessment is essential for maintaining roadway safety and optimizing maintenance investments. However, conventional assessment methods such as manual visual inspections and specialized sensing equipment are often time-consuming, expensive, and difficult to scale across large networks. Recent advancements in generative artificial intelligence (GAI) have introduced new opportunities for automating visual interpretation tasks using street-level imagery. This study evaluates the performance of seven multimodal large language models (MLLMs) for road surface condition assessment, including three proprietary models (Gemini 2.5 Pro, OpenAI o1, and GPT-4o) and four open-source models (Gemma 3, Llama 3.2, LLaVA v1.6 Mistral, and LLaVA v1.6 Vicuna). The models were tested across four task categories relevant to pavement management: distress and feature identification, spatial pattern recognition, severity evaluation, and maintenance interval estimation. Model performance was assessed across five dimensions: response rate, response correctness, consistency, multimodal errors, and overall computational intensity and cost. Results indicate that MLLMs can interpret street-level imagery and generate task-relevant outputs in a cost-effective manner. Among the evaluated models, we recommend GPT-4o as the preferred option, as it balances responsiveness, accuracy, and computational cost.

Runway safety in landing and takeoff operations is strongly influenced by skid resistance, which may be reduced by rubber buildup from aircraft braking. Conventional friction measurement methods require specialized equipment and operational interruptions, limiting their frequency and increasing costs. This study explores the use of satellite imagery and convolutional neural networks (CNNs) as a non-intrusive and cost-effective alternative for classifying runway friction coefficients. A dataset of about 7,000 Google Earth images of Brazilian aerodrome runways was matched with measured friction coefficients. Three preprocessing techniques (contrast limited adaptive histogram equalization [CLAHE], Gaussian filter, and wavelet) and five CNN architectures (simple CNN, ResNet50, DenseNet121, InceptionV3, and VGG16) were evaluated. The best performance was obtained with CLAHE + Gaussian preprocessing and DenseNet121, achieving over 74% accuracy in cross-validation. External validation with unseen satellite images confirmed robustness, while tests with unmanned aerial vehicle imagery showed limited generalization across visual domains. The results indicate that deep learning combined with remote sensing can support airport pavement management. The proposed approach complements direct friction monitoring, offering a scalable, low-cost, and non-intrusive tool to identify critical areas, guide preventive maintenance, and enhance operational safety.

This study proposes an intelligent back-calculation framework to estimate multilayer pavement elastic moduli from FWD deflection data under realistic measurement uncertainty. A spectral element method (SEM) model is used to simulate transient FWD responses and generate large-scale datasets. A Transformer regression model is trained to map peak deflection basins to layer moduli, considering four noise scenarios (no error, random, systematic, and combined). Baseline models (BPNN, SVR, and XGBoost) are also evaluated for comparison. The proposed SEM–Transformer framework achieves strong accuracy and robustness, with average R 2 >0.94 and MAPE < 8% across all noise cases, and shows superior performance for the base course under noisy conditions. The results demonstrate a reliable and efficient data-driven feasibility framework to support pavement structural evaluation and future digital-twin-based pavement management.

A reliable decision model in pavement management systems (PMS) is important to maintain the quality and serviceability of pavement systems in the face of increasing traffic volumes and more severe climate conditions. Even though accurate pavement condition assessment is essential for a decision model, there are limited approaches to practically evaluate the structural condition of concrete slabs and the functional condition of concrete pavements. Thus, this study identified practical methods to estimate the remaining service life (RSL) for pavement condition evaluation and developed a decision framework to recommend more appropriate maintenance strategies for concrete pavements. The stress-to-strength ratio (SSR) derived from falling weight deflectometer (FWD) deflection data was selected to estimate RSL based on the structural capacity of concrete slabs. The international roughness index (IRI) was used to estimate RSL based on functional conditions, and a new IRI prediction model was developed for concrete pavements. These structural and functional condition-based RSLs were integrated into the proposed decision framework to provide a more reliable basis for maintenance decisions. Since the proposed framework requires only FWD and IRI data, the most common non-destructive test data sets, it can be practically incorporated into current PMS practices. Furthermore, the RSL calculation equations can be easily calibrated for local conditions, and SSR and IRI threshold values can be adjusted to meet agency needs. This flexibility allows the proposed framework to be implemented by other agencies or in different countries. Therefore, the proposed decision framework can potentially improve current PMS practices, leading to better quality of concrete pavements.

This paper conducts an investigation into the application of Markov processes for optimizing the lifecycle management of pavements on the Irish national road network. Transition probability matrices (TPMs) are developed for a range of condition indicators, incorporating additional factors such as annual average daily traffic. The influence of these factors on the derived TPMs is illustrated. A methodology for updating the TPMs to facilitate consideration of the effects of incorporation of new materials, environmental management and/or of the effects of climate change and advance deterioration is considered. The research assesses the merits of employing homogeneous versus non-homogeneous Markov chains. The findings underscore the significant impact of probabilistic techniques in the domain of pavement asset management but considering budget allocation across the network for a specified time horizon.

The volumetric mix design methods are not sufficient to address the complex nature of today’s asphalt mixtures. Hence, it is essential to incorporate performance tests in the asphalt mix design process. Integrating these tests requires the development of asphalt mixtures performance thresholds. In this study, five construction projects were selected for field pilot sections to test the performance thresholds, and the proposed Balanced Mix Design (BMD) method developed for Oregon. The objective is to compare the performance of asphalt pavements designed using volumetric, and BMD approaches in the laboratory and field. Testing the production mixes revealed that four of the five projects designed with the BMD exhibited superior cracking performance than volumetric mixes. Despite all mix designs passing the cracking thresholds in the laboratory, four projects failed to meet these standards with plant-produced mixes. Nevertheless, pavement management system data revealed the similarity in the pavement segments produced with volumetric and BMD approaches in terms of rut depth and surface roughness. Notably, the rutting resistance of all the BMD mixes was excessive, necessitating a need to increase the BMD rut depth thresholds to soften the asphalt mix and improve the cracking resistance of the asphalt mixtures in Oregon, United States.

Computational optimization for low-carbon and circular pavement management at the network level

Accurate imputation of missing pavement-condition data is critical for proactive infrastructure management, yet it is complicated by spatial non-stationarity—deterioration patterns and data quality vary markedly across regions. This study proposes a Spatially Gated Mixture-of-Experts (SG-MoE) imputation model that explicitly encodes spatial heterogeneity by (i) clustering road segments using geographic coordinates and (ii) supervising a gating network to route each sample to region-specialized expert regressors. Using a large-scale national pavement management database, we benchmark SG-MoE against a strong baseline under controlled missingness mechanisms (MCAR: missing completely at random; MAR: missing at random; MNAR: missing not at random) and missing rates (10–50%). Across scenarios, SG-MoE consistently matches or improves upon the baseline; the largest gains occur under MCAR and the challenging MNAR setting, where spatial specialization reduces systematic underestimation of high crack-rate sections. The results provide practical guidance on when spatially aware ensembling is most beneficial for infrastructure imputation at scale. We additionally report comparative results under three missingness mechanisms. Across five random seeds, SG-MoE is comparable to the single LightGBM baseline under MCAR/MAR and achieves its largest gains under MNAR (e.g., sMAPE improves by 0.82 points at 10% MNAR missingness).

Assessing pavement conditions in Nepal is costly and time-consuming, with rising traffic and aging infrastructure making maintenance increasingly challenging. This study developed and compared pavement deterioration models to predict the Surface Distress Index (SDI) without manual assessment, using historical road data. SDI was modeled as a function of five key factors: International Roughness Index (IRI), pavement age, total annual rainfall, annual temperature range, and commercial vehicle traffic. Data were collected from relevant government sources, covering 157 road sections with a combined length of 15,783 km, for the period from 2012 to 2022. Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models were developed for SDI prediction. MLR analysis, conducted in Microsoft Excel, assessed statistical significance through ANOVA, R² values, and regression coefficients. In contrast, ANN modeling utilized a Multi-Layer Perceptron (MLP) architecture implemented in TensorFlow and Keras. The ANN model was optimized through iterative experimentation with varied architectures, employing ReLU activation and the Adam optimizer for adaptive learning. The study evaluated a range of architectures, beginning with simple single-layer networks and extending to Deep Neural Networks (DNNs) with up to four hidden layers. Results showed that, during model development, MLR achieved an R² of 0.735, whereas the ANN model, with a 5-232-1 structure and 104 epochs, outperformed MLR with an R² of 0.809. Validation of both models indicated strong alignment between observed and predicted values, with ANN demonstrating superior predictive accuracy (R² = 0.816) compared to MLR (R² = 0.74). The error histogram further confirmed ANN’s better performance, which confirms its improved reliability. The study highlights the effectiveness of both models while emphasizing ANN’s advantage in capturing complex nonlinear relationships. These findings suggest that integrating ANN into Nepal’s pavement management framework can enhance predictive accuracy, reduce assessment costs, and support more efficient maintenance planning.

ABSTRACT Automated pavement management systems are essential for detecting, and assessing road cracks to provide accurate, timely data for maintenance planning and resource allocation. Traditional methods rely on large labeled datasets to capture diverse crack patterns, while semi-supervised methods, although beneficial, still require substantial amounts of unlabeled data. This dependency poses challenges in real-world applications where both labeled and unlabeled data can be scarce. In such low-data environments, models may struggle to detect faint crack edges and generalize well, which can lead to overfitting and decreased accuracy. Furthermore, accurately capturing crack width variations, critical for detecting irregular widening, remains a challenge. To address these limitations, we propose a teacher-student framework for crack segmentation and severity assessment designed to work effectively with limited labeled data. This framework precisely measures crack width, type, spacing, and severity, which are essential metrics for Pavement Crack Severity Index (PCSI) evaluation and GPS-based mapping. Testing on real-world datasets and through two application studies reveals that our framework outperforms six other models in segmentation accuracy with limited labeled data. Additionally, it accelerates the mapping process by approximately 76.27% compared to conventional methods, providing an efficient, data-driven solution for road pavement monitoring and maintenance.

Geothermal energy offers a promising and sustainable approach for maintaining and improving road infrastructure, particularly in mitigating ice formation and enhancing thermal performance. This review systematically examines scientific studies addressing the application of geothermal systems for the thermal management of road pavements. A structured methodological framework, following PRISMA guidelines, was employed to identify and select relevant publications from official scientific databases, focusing on articles considered eligible based on their thematic relevance and practical application within the field. The review pursues three complementary objectives: (i) to provide a comprehensive synthesis of the current literature, (ii) to analyze research trends, including modeling strategies, laboratory experiments, and field applications, and (iii) to evaluate reported system performance in terms of efficiency, design parameters, and environmental considerations. Contributions are categorized into simulation-based studies, experimental investigations, and combined approaches, allowing for comparison across methodologies and climatic contexts. Key findings, technological limitations, and challenges encountered in the literature are discussed, including system efficiency, design constraints, and environmental considerations. By synthesizing the current state of knowledge, this review highlights critical gaps and potential avenues for future research, offering guidance for the development of innovative geothermal-based solutions. The insights presented herein contribute to informed decision-making in research planning and infrastructure development, supporting safer, more energy-efficient, and sustainable road systems.

Recent advances in vibration-based pavement assessment have enabled the low-cost monitoring of road conditions using inertial sensors and machine learning models. However, most studies focus on isolated tasks, such as roughness classification, without integrating statistical validation, anomaly detection, or maintenance prioritization. This study presents a unified framework for road roughness severity classification and predictive maintenance using multi-axis accelerometer data collected from urban road networks in Pretoria, South Africa. The proposed pipeline integrates ISO-referenced labeling, ensemble and deep classifiers (Random Forest, XGBoost, MLP, and 1D-CNN), McNemar’s test for model agreement validation, feature importance interpretation, and GIS-based anomaly mapping. Stratified cross-validation and hyperparameter tuning ensured robust generalization, with accuracies exceeding 99%. Statistical outlier detection enabled the early identification of deteriorated segments, supporting proactive maintenance planning. The results confirm that vertical acceleration (accel_z) is the most discriminative signal for roughness severity, validating the feasibility of lightweight single-axis sensing. The study concludes that combining supervised learning with statistical anomaly detection can provide an intelligent, scalable, and cost-effective foundation for municipal pavement management systems. The modular design further supports integration with Internet-of-Things (IoT) telematics platforms for near-real-time road condition monitoring and sustainable transport asset management.

In the context of pavement management systems (PMSs), overloaded trucks impose severe economic and environmental burdens by accelerating pavement deterioration and increasing greenhouse gas (GHG) emissions. Existing research on Weigh-in-Motion (WIM) placement has rarely incorporated environmental impacts, particularly greenhouse gas (GHG) emissions, into the decision-making process. Instead, most studies have focused on infrastructure damage and have paid limited attention to how enforcement interacts with driver evasion behavior and schedule-related constraints. To address this gap, this study develops a bi-level optimization framework that simultaneously minimizes PMS costs, travel costs, and environmental (GHG) costs. The upper-level problem represents the total social cost minimization, while the lower-level problem models drivers’ routes and demand shift. The framework endogenously captures utility-based demand shifts, allowing overloaded drivers to switch to legal operations when enforcement and schedule-related constraints outweigh overloading benefits. A numerical study using the Sioux Falls network demonstrates that dual WIM installations significantly outperform single configurations, achieving network-wide cost reductions of up to 1.5% compared to 0.4%. Notably, PMS costs for overloaded trucks decreased by nearly 60%, confirming the effectiveness of strategic enforcement. Ultimately, this study contributes a unified decision-support tool that reframes WIM enforcement from a passive control measure into a proactive strategy for sustainable freight management.

Airport pavement condition assessment plays a critical role in ensuring operational safety, surface functionality, and long-term infrastructure sustainability. Traditional visual inspection methods, although widely used, are increasingly challenged by limitations in accuracy, subjectivity, and scalability. In response, the field has seen a growing adoption of automated and intelligent inspection technologies, incorporating tools such as unmanned aerial vehicles (UAVs), Laser Crack Measurement Systems (LCMS), and machine learning algorithms. This systematic review aims to identify, categorize, and analyze the main technological approaches applied to functional pavement inspections, with a particular focus on surface distress detection. The study examines data collection techniques, processing methods, and validation procedures used in assessing both flexible and rigid airport pavements. Special emphasis is placed on the precision, applicability, and robustness of automated systems in comparison to traditional approaches. The reviewed literature reveals a consistent trend toward greater accuracy and efficiency in systems that integrate deep learning, photogrammetry, and predictive modeling. However, the absence of standardized validation protocols and statistically robust datasets continues to hinder comparability and broader implementation. By mapping existing technologies, identifying methodological gaps, and proposing strategic research directions, this review provides a comprehensive foundation for the development of scalable, data-driven airport pavement management systems.

Incorporating the structural condition of pavements into the maintenance and rehabilitation treatment selection process is essential for improving performance and prolonging service life. However, because of challenges in network-level structural condition testing, agencies often rely solely on surface-based condition data. This paper proposes and evaluates an approach to incorporate structural condition information from a traffic speed deflectometer into the pavement management decisions of the Virginia Department of Transportation in the U.S. The approach follows the AASHTO design procedure to calculate the effective structural number (SN eff ) and uses SN eff to determine the remaining structural life (RSTL). Analysis of 4,250 lane-miles of interstate and primary roads revealed that, while 90%, 80%, and 30% of primary roads had an RSTL of at least 5, 10, and 20 years, respectively, interstate roads were in better condition, with 91%, 88%, and 82% having an RSTL of at least 5, 10, and 20 years, respectively. A structural condition indicator based on RSTL thresholds for selecting treatment categories was proposed. Final treatment recommendations were derived by combining the current surface-distress-based categories with the proposed structural-condition-based categories. Unconstrained needs analysis, including detailed case studies, showed differences in needs when structural condition was considered, compared with using only surface-based condition information. These differences in needs varied with the existing pavement condition. While the proposed RSTL approach can enhance treatment selection, a systematic investigation is recommended to determine appropriate intervals between structural testing cycles at the network level.

Accurate modeling of pavement performance is vital to maintaining safe, reliable, and sustainable transportation infrastructure. This review synthesizes current approaches to pavement deterioration modeling, with emphasis on key influencing factors, performance indicators, and methodologies employed within Pavement Management Systems (PMS). Primary deterioration drivers, including traffic loading and environmental stressors, are analyzed for their impact on degradation patterns. Performance indicators such as the Pavement Surface Evaluation and Rating (PASER), Pavement Condition Index (PCI), and International Roughness Index (IRI) are evaluated for their effectiveness in capturing pavement condition and guiding maintenance decisions. Modeling techniques are broadly categorized into deterministic, probabilistic, and intelligent (machine learning–based) frameworks to illustrate the evolution of predictive approaches. Across these approaches, the notion of symmetry can be interpreted as the balance and consistency achieved between model assumptions, input variables, and predicted pavement behavior, while asymmetry represents deviations caused by uncertainty, variability, and nonlinearity inherent in real-world conditions. Recognizing these symmetrical and asymmetrical relationships helps unify different modeling paradigms and provides insight into how each framework handles equilibrium between accuracy, complexity, and interpretability. The review also highlights persistent challenges in data availability, quality, and standardization. Notably, the increasing adoption of machine learning reflects its capacity to handle high-dimensional and spatiotemporal datasets. Recommendations are proposed to improve the robustness, scalability, and transparency of future deterioration models, thereby enhancing their role in data-driven, resilient, and cost-effective pavement management strategies.

Thin asphalt overlays are widely used in Wyoming and have various impacts on pavement performance. This study evaluates the effectiveness of thin overlay treatments applied on interstates and non-interstates between 2010 to 2017, using the Pavement serviceability index (PSI) as a performance indicator. Historical PSI values were obtained from the pavement management system database to assess the performance of treatments. A deterministic methodology was introduced to quantify the effectiveness of thin overlays, followed by a survival analysis to analyze the data and predict the survival characteristics of the treatments. The results indicate that the service life of thin asphalt overlays ranges from seven to eight years based on PSI. Treated bases appear to enhance long-term overlay performance, while traffic significantly affects treatment performance. The findings of this study can contribute to improved design and strategy development for overlay treatments in Wyoming.

Abstract Road safety is strongly influenced by pavement friction, which governs tire traction and braking efficiency, especially in wet conditions. Conventional friction evaluation methods, while widely used, are time-consuming, costly, and often lack generalizability. This study investigates advanced machine learning (ML) techniques for predicting the friction number of continuously reinforced concrete pavement (CRCP) using data from the Long-Term Pavement Performance (LTPP) database. The dataset comprised 170 observations from 33 CRCP sections, representing different climatic and structural conditions. A 5-fold cross-validation approach was employed within MATLAB’s Regression Learner App to ensure robust and unbiased model evaluation. Six ML models were examined, including regression trees, support vector machines (SVM), ensemble methods, Gaussian Process Regression (GPR), artificial neural networks (ANN), and kernel-based approaches. Results show that the Rational Quadratic GPR model achieved the highest predictive accuracy (R² = 0.70, RMSE = 5.29, MAE = 3.90), outperforming other conventional machine learning algorithms used for comparison. Feature importance and sensitivity analyses revealed that pavement age, traffic loads, thickness, temperature, and humidity are the most influential factors affecting surface friction. The findings provide practical insights for data-driven pavement management, offering transportation agencies reliable tools to enhance safety, optimize maintenance strategies, and extend pavement service life. Although the dataset size is moderate, the consistent cross-validation results indicate strong model reliability; future studies using larger and more diverse datasets are recommended to further validate the model’s generalizability.

Purpose The purpose of this study is to develop a predictive tool that enables urban facilities managers and transportation authorities to forecast transverse cracking in Continuously Reinforced Concrete Pavement (CRCP) with high accuracy. By integrating Genetic Algorithm (GA) with Support Vector Regression (SVR), this study seeks to enhance pavement maintenance planning, reduce infrastructure deterioration and support sustainable asset management practices in urban environments. This study contributes to data-driven decision-making aligned with Sustainable Development Goals (SDGs), particularly SDG 9 (industry, innovation and infrastructure) and SDG 11 (sustainable cities and communities). Design/methodology/approach This study uses a hybrid machine learning model combining SVR with GA optimization to predict transverse cracking in CRCP. A data set comprising 395 records from 33 pavement sections in the Long-Term Pavement Performance database is used. Predictor variables include structural, climatic and traffic-related features. The model’s performance is evaluated using five-fold cross-validation and benchmarked against traditional models such as Linear Regression and Decision Trees to assess accuracy and reliability in forecasting pavement distress. Findings The GA-SVR model achieved a mean root mean square error of 4.37 and an R² of 0.921, significantly outperforming Linear Regression and Decision Tree models. Variables such as pavement age, total thickness, precipitation and initial roughness were identified as key predictors. Three-dimensional (3D) interaction plots revealed how these factors jointly influence pavement performance over time. Residual analysis confirmed the robustness of the model, showing randomly distributed errors without systematic bias. These results demonstrate that GA-SVR is a reliable and accurate tool for forecasting pavement cracking in urban infrastructure systems. Originality/value To the best of the authors’ knowledge, this study is among the first to apply GA-optimized SVR for predicting transverse cracking in CRCP within the context of urban facilities management. This study bridges a gap between pavement engineering and smart infrastructure planning by integrating machine learning with sustainable development objectives. The model’s high accuracy and practical utility provide a novel contribution to proactive urban asset management, aligning technological innovation with citizen-focused outcomes. Its methodological framework can be adapted to other infrastructure systems, offering a scalable solution for smart city initiatives.