This study proposes a novel approach to driver drowsiness detection using the Video Vision Transformer (ViViT) model, which captures both spatial and temporal dynamics simultaneously to analyze eye conditions and head movements. The National Tsing Hua University Driver Drowsiness Detection (NTHU-DDD) dataset, which consists of 36,000 annotated video clips, was utilized for both training and evaluation. The ViViT model is compared to traditional Convolutional Neural Network (CNN) and Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models, demonstrating superior performance with 96.2% accuracy and 95.9% F1-Score, while maintaining a 28.9 ms/frame inference time suitable for real-time deployment. The ablation study indicates that integrating spatial and temporal attention yields a notable improvement in model accuracy. Furthermore, positional encoding proves essential in preserving spatial coherence within video-based inputs. The model’s resilience was tested across a range of challenging conditions including low-light settings, partial occlusions, and drastic head movements and it consistently maintained reliable performance. With a compact footprint of just 89 MB, the ViViT model has been fine-tuned for deployment on embedded platforms such as the Jetson Nano, making it well-suited for edge AI applications. These findings highlight ViViT’s promise as a practical and high-performing solution for real-time driver drowsiness detection in real-world scenarios.
Traffic congestion in urban commercial districts presents a critical challenge to sustainable mobility, particularly in developing cities. This study addresses this issue by developing and calibrating a mesoscopic simulation model to optimize traffic performance parameters in the commercial district of Ayacucho, Peru. The methodology was based on extensive fieldwork to gather traffic volume, travel time, and parking data. Using this data, a PTV Vissim model was developed and rigorously calibrated, with its accuracy validated through the Geoffrey E. Havers (GEH) statistic. Various traffic management strategies, including signal timing adjustments and parking supply regulation, were simulated and evaluated. The results indicate a substantial improvement in network performance: Average intersection delay was reduced from 10.72 seconds to 7.40 seconds, and a significant decrease in queue lengths was observed. The findings confirm that calibrated mesoscopic simulation serves as a robust and effective tool for quantitatively assessing traffic interventions, thereby providing municipal authorities with reliable data for evidence-based urban planning.
Pavement distress is a critical factor in road maintenance planning, directly influencing transportation safety, serviceability, and infrastructure costs. While traditional mechanistic and statistical models provide limited accuracy, they often fail to capture the nonlinear and multi-factorial nature of pavement deterioration. This study addresses this gap by proposing an integrated machine learning (ML) framework that incorporates real-time traffic and climatic variables for predicting pavement roughness. The framework draws on multiple open-source datasets, Long-Term Pavement Performance (LTPP), Federal Highway Administration (FHWA) traffic volumes, and National Oceanic and Atmospheric Administration (NOAA) climate records, to construct a multidimensional feature space. Four predictive algorithms were benchmarked: Random Forest (RF), XGBoost (XGB), Support Vector Machine (SVM), and Multi-layer Perceptron (MLP). Ensemble-based models achieved superior predictive accuracy, with Random Forest attaining R$^2 \approx$ 0.89 and Root Mean Square Error (RMSE) $\approx$ 0.61, outperforming traditional regression baselines. The findings highlight that ensemble learning can more effectively capture non-linear dependencies between structural, traffic, and climatic factors than alternative approaches. Beyond technical performance, the study illustrates the potential of integrating continuously updated environmental and traffic data into pavement management systems, offering a pathway to more cost-efficient, reliable, and sustainable maintenance planning.
Aceh Province is a critical case for freight and infrastructure studies due to its geographic isolation, post-disaster recovery context, and heavy dependence on roads for over 95% of commodity transport. Despite its rich agricultural output, limited multimodal infrastructure hampers efficient distribution. This study aims to (1) analyze the effect of road network connectivity on commodity transportation and regional development, and (2) develop a forecasting model to predict future commodity transportation needs in Aceh Province. The Structural Equation Modeling (SEM) was applied to analyze the relationships among Road Network Connectivity (RNC), Freight Transport (FT), and Regional Development (RD), using data from 400 respondents across 23 districts. The SEM results show all latent variables are interconnected. FT plays a strong mediating role, linking connectivity improvements to development benefits. The study also develops forecasting models for commodity generation and attraction based on population, expressed as $Y$ = 2.209 $X_1$ and $Y$ = 2.807 $X_1$. These models highlight population as a reliable predictor of freight demand and can be generalized to other regions with similar geographic and infrastructure constraints. This research introduces a novel SEM-based framework for freight analysis in Indonesia and offers policy insights for integrating road infrastructure planning with regional development strategies.
Effective traffic management at signalized intersections is crucial for enhancing fuel efficiency, safety, and mobility; however, this is challenging for human drivers due to a lack of predictability. This paper proposes a predictive vehicle control system that extends a traditional human-based driving model to optimize traffic flow, reduce intersection transit time, and fuel consumption. The proposed system utilizes an optimal trajectory prediction model to determine the stopping velocity pattern at traffic signals and employs safety gap synchronization, thereby exhibiting human-like car-following behavior. Specifically, the optimal velocity profiles are generated based on a trajectory optimization model over a long-time horizon. A polynomial function is fitted with these optimal trajectories to find the ideal stopping pattern. Instead of repeating the optimization at each step, as in the Model Predictive Control (MPC) approach, our method determines the control acceleration with necessary adjustments while ensuring driving safety. Moreover, the synchronization compensation factor improves the transition from idling to driving conditions. Performance evaluation through microscopic traffic simulations demonstrates improvements in intersection throughput and fuel efficiency, showcasing the effectiveness of the proposed predictive vehicle control system. Unlike the computationally demanding MPC approach, our proposed system offers a practical balance between real-time applicability and traffic flow efficiency.
Maritime transportation safety is a strategic priority in Indonesia’s national logistics system as an archipelagic country. However, high shipping intensity and limited inspection resources pose serious challenges to the effectiveness of ship inspections. Ship Safety Inspectors (PPKK) play a crucial role in ensuring the seaworthiness of ships through inspections of nautical, technical, and communication (radio) aspects; however, their functional contributions have rarely been systematically examined. This study employs a quantitative approach using survey methods and Structural Equation Modeling–Partial Least Squares (SEM-PLS) analysis techniques. A sample of 64 CSOs from three strategic ports (Tanjung Priok, Soekarno-Hatta, Makassar, and Sorong-Pelra) was analyzed to examine the relationship between inspection functions, workload, and maritime safety. The results indicate that technical, nautical, and communication functions significantly influence workload, while technical functions and workload have a direct impact on maritime safety. Work volume acts as a mediating variable in the model. The implications of these findings emphasize the importance of enhancing the competence of PPKK and developing risk-based inspection policies. This study provides empirical contributions to the reform of the maritime safety supervision system and opens opportunities for cross-national validation to strengthen the generalizability of the model.
A Traffic Impact Assessment (TIA) is crucial in urban and transportation planning, especially in densely populated areas. Key components of a TIA include trip generation, trip distribution, modal split, and assignment. Among these, trip generation forecasting is the most significant because it influences land-use decisions and supports sustainable transportation strategies. This study conducted a Systematic Literature Review (SLR) following the Reporting Standards for Systematic Evidence Syntheses (ROSES) protocol. A total of 21 peer-reviewed articles were selected from the Scopus and Web of Science databases. The review focused on how machine learning (ML) techniques are used to enhance the accuracy of trip generation. Thematic analysis revealed five main themes: prediction model development, urban planning decisions, urban sustainability, forecasting challenges, and innovative ML applications. Standard models include Artificial Neural Networks (ANNs), Support Vector Machines, and Random Forests. Incorporating ML in trip generation forecasts improves the accuracy and reliability of TIA processes. These techniques help identify key variables that affect travel behavior, supporting more effective and sustainable urban transportation planning and decision-making.
Efficient light rail transit (LRT) systems are crucial for sustainable urban mobility; however, unforeseen departure delays continue to be a major hurdle, undermining operational reliability and passenger satisfaction. This study establishes a data-driven framework for forecasting departure delays by combining static GTFS schedules with real-time GTFS operational data from the Canberra LRT system. The dataset included 15,538 records with 42 attributes, spanning from 28 August 2020 to 13 August 2022. A stringent preprocessing pipeline was implemented, encompassing temporal feature engineering and feature selection based on mutual information. The Random Forest regressor with feature engineering and selection (RFR-FEFS) attained the highest predictive performance on the test set ($R^2$ = 0.94, MAE = 2.93, MSE = 34.32). The high accuracy indicates the model’s efficacy, yet it necessitates careful evaluation of potential overfitting and its generalizability beyond the examined system. Ablation experiments were performed to assess the impact of various feature groups by omitting temporal, spatial, or operational attributes. The findings indicate that the exclusion of temporal features decreased $R^2$ to 0.90, the exclusion of spatial features reduced it to 0.93, and the exclusion of operational features resulted in the most significant decline to 0.23. These findings affirm that all three feature categories contribute distinctly and synergistically to model performance. This research illustrates the capability of integrating diverse GTFS data with sophisticated machine learning techniques to attain precise LRT delay forecasts. Nevertheless, the framework was validated solely on one system and time frame; future research should investigate its transferability to other cities and integrate supplementary contextual data, including meteorological conditions and incident reports, to improve robustness and practical applicability.
Continuous improvement in service quality assurance, based on customer satisfaction, is critical for loading and unloading activities at dry bulk ports. Many ports are now adopting and refining various methods in response to the advancements of Industry 4.0. This research aims to develop and implement Adaptive DMAIC 4.0. Key advantages of this method include IoT based real-time monitoring systems, predictive data analytics, and process automation capabilities. Current Six Sigma measurements show level 3 (DPMO 11,800). While the Cp value of 1.19 indicates stable process stability, the Cpk value of 0.76 $<$ 1 reveals remaining issues requiring systematic, continuous improvement. To enhance process performance, the average loading/unloading time should be maintained closer to the target midpoint of 1.5 minutes/bulk, creating a more balanced distribution. This adjustment would help increase the Cpk value to meet the minimum standard $\geq$ of 1.33, ensuring consistently efficient operations. In theory, implementing the DMAIC 4.0 framework will establish a system that is more resilient to internal and external disruptions, enables sustained performance improvement, and drives toward zero defects and Six Sigma capability. In practice, this approach significantly enhances loading and unloading performance for boosting capacity, operational capability, and TKBM professionalism while eliminating human error.
This study develops and validates an integrated model for evaluating passenger service quality (SQ) in Thailand’s rural railway system by embedding environmental and engineering perspectives within the RAILQUAL framework. Drawing upon SERVQUAL, Grönroos's model, and Servicescape theory, it introduces the Eco-Rail Atmosphere Quality (Eco-RAQ) construct, which incorporates sustainability attributes-greenhouse-gas reduction, waste management, traction-energy efficiency, and renewable energy efficiently-into the Rail Atmosphere Quality (RAQ) dimension. Survey data from 1,013 passengers were analyzed using covariance-based structural equation modeling (CBSEM). The final model exhibits excellent fit ($\chi^2$/df = 1.096, CFI = 1.000, RMSEA = 0.010) and explains 91.5% of variance in rail efficiency quality (REQ. RAQ demonstrates the strongest total effect on REQ ($\beta$ = 0.848, $p$ $<$ 0.001), while Eco-RAQ shows a meaningful but more modest total effect ($\beta$ = 0.257, $p$ $<$ 0.001), influencing REQ both directly and indirectly through rail perceived quality (RPQ). Validity diagnostics confirm discriminant validity (HTMT $<$ 0.85) and no substantive common-method bias. The findings advance service-quality theory by integrating sustainability cognition into the Stimulus-Organism-Response paradigm and by proposing Eco-RAQ as a socio-technical mechanism linking passenger perception with operational performance. The model offers actionable insights for achieving Sustainable Development Goals (SDGs) 9, 11, and 13 in rural rail contexts.
Recognition of traffic signs by drivers is essential for ensuring road safety. Recently, with the growing demand for driver assistance systems and autonomous vehicles, traffic sign recognition has become increasingly important. In this study, Spatial Transformer Networks (STN) integrated with Convolutional Neural Networks (CNN) were used to classify traffic signs. STNs minimize the effects of geometric distortions by applying affine transformations to images, thereby improving classification performance. This study focuses on adapting and optimizing an STN-based CNN model specifically for the Russian Traffic Signs Dataset (RTSD) to achieve higher classification accuracy. The proposed model was trained and tested on the RTSD. First, the proposed CNN model was trained on the RTSD-R1 and RTSD-R3 datasets, achieving accuracy rates of 89.15% and 94.3%, respectively. Then, by integrating STN into the CNN model, the proposed model was trained on the RTSD-R1 and RTSD-R3 datasets, achieving accuracy rates of 93% and 95%, respectively. These results demonstrate that incorporating STNs into the CNNs is effective in improving traffic sign classification performance.
The transition toward sustainable port management has intensified interest in how institutional pressures and organizational capabilities shape environmental and operational outcomes. This study investigates how environmental regulation, stakeholder pressure, employee training, and managerial commitment influence green engineering infrastructure and innovation and, in turn, green port implementation. The model specifies a serial mediation in which green engineering infrastructure and innovation and green port implementation connect institutional drivers to environmental performance and operational efficiency. Survey data from 221 respondents in two Indonesian container terminals were analyzed using partial least squares modeling. Results show that environmental regulation is the most influential driver, while stakeholder pressure, training, and managerial commitment reinforce capability building and adoption of low emission technologies. Green engineering infrastructure and innovation facilitates green port implementation, which significantly improves environmental performance and operational efficiency. Theoretically, the study extends institutional and resource based perspectives by clarifying how two stage mediation translates institutional pressures into dual sustainability outcomes in port settings. Practically, the findings show that sustainable port transformation in emerging economies depends on aligning regulation with investments in human capital and green technologies, guiding policymakers and port authorities.
The purpose of the research was to find out the most efficient method to measure the longitudinal profile irregularities of the railway track geometry. Some researchers used an accelerometer installed at the axle-box bogie to monitor the longitudinal profile of the railway track. However, the method was risky because the accelerometer could disappear due to improper installation. In this research, the accelerometer was installed in the train cabin. The scope of analysis started with the IMU calibration by manual technique, filtering the Z-axis acceleration of IMU data by the Kalman Filter method, and converting to longitudinal profile data by double integral calculation. The normal distribution of the longitudinal profile of the railway track based on the acceleration data and measurement data of the track geometry measuring train (TGMT) was tested by the Kolmogorov-Smirnov Method. The paired comparison test used the Wilcoxon signed-ranks test method, and test results showed the P-value is 0.725 (left rail) and 0.073 (right rail), which are greater than 0.05. Therefore, no difference between the longitudinal profile based on the TGMT data and the longitudinal profile based on accelerometer data analysis, and the longitudinal profile irregularities can be monitored by using the accelerometer in the train cabin. The use of this method will support SDGs 9, build resilient infrastructure.
Mass Rapid Transit (MRT) systems play a critical role in promoting sustainable development, particularly in megacities. This study assesses the quality of boarding/alighting facilities along with accessibility of MRT, as integrated system components which is vital for maintaining a safe, efficient and user-friendly transit system, in the context of built-up cities. A robust questionnaire form is designed using 29 selected variables derived from pilot survey which was administerd to 1,397 respondents across nine operational stations of MRT in Dhaka, a developing megacity of Southeast Asia. Using the collected data, Gini Index and ANOVA are employed for variable prioritization. Machine Learning Algorithms, i.e., Random Forest (RF) Classifier, Support Vector Machine (SVM) and Classification and Regression Trees (CART), are compared to assess predictive performance where RF demonstrated better performance based on accuracy. Additionally, feature selection identified critical factors related to MRT trip performance, such as switching cost comparison, feeder service cost, inclusive service performance, customer loyalty, lighting near stations, overall comfort, security. This study, further, incorporates two most crucial factor, switching cost comparison and feeder service cost to a hybrid function, assessing system components and user transferability, utilizing a novel matrix-based approach. The study’s conclusions provide insights into boarding/alighting facility and accessibility as system components incorporating hybrid cost function (HCF) to enhance the efficiency of MRT services in built-up cities across the world.
This study investigates commuter mode choice behavior between Bulacan and Quezon City, in the context of the forthcoming Metro Rail Transit Line-7 (MRT7). In this study we examine current travel behavior and preferences using both revealed and stated preference (SP) data, while also describing the travel characteristics of private car users and the operational features of other public transportation services. We hypothesize that in-vehicle travel time and out-of-pocket travel cost are the most influential factors affecting the mode choice decisions. A total of 4600 survey responses were collected using orthogonally designed choice set and analyzed using Discrete Choice Modeling (DCM) through Multinomial Logit Framework (MNL) in LIMDEP NLOGIT and RStudio. Modal shift analysis shows in-vehicle travel time, cost relative-to-income, access-time, comfort and safety as the most influential factors shaping mode choice. More commuters choose rail-based transportation when in-vehicle travel time reaches 62.54 minutes. When it comes to out-of-pocket cost, a modal shift occurs around a cost-to-income ratio above 8% (equivalent to 120.87 pesos). The average access-time for road-based transportation is 22.9 minutes. Though less influential, a notable shift is observed at above 30 minutes access-time. Comfort shows a strong behavioral influence with rail commuters rising from 34.62% under poor comfort to 80.86% under high comfort. Similarly, safety perception affects choice: rail captures 77.57% when road modes are viewed as unsafe, but only 35.16% when they are seen as safe. Overall, in-vehicle travel time and fare affordability demonstrated the highest sensitivity in determining mode choice, indicating that commuters prioritize the efficiency of their travel time and their financial capacity to afford the fare.
Traditional road maintenance strategies often focus solely on detecting cracks, neglecting the structural complexity crucial for prioritizing repairs. This study introduces a computational framework that combines deep learning-based segmentation with graph-theoretic analysis to automatically quantify critical topological features of crack networks, such as branching points and closed loops. Three segmentation models—DeepLabV3, Attention U-Net, and SegFormer—are evaluated on the newly developed Timor-Leste Crack (TLCrack) dataset and the publicly available CrackForest benchmark, leveraging topology-aware loss functions and evaluation metrics. The resulting segmentation outputs are skeletonized and converted into graph structures, enabling automated measurements of branch points and cyclic regions. Experimental findings reveal that Attention U-Net achieves the highest topological accuracy, with a Betti-0 error of 1.70 $\pm$ 0.62 on the TLCrack dataset. Additionally, the graph-based quantification module demonstrates robust performance, achieving a branch point counting mean absolute percentage error (MAPE) of 5.33% and flawless closed-loop detection on the same dataset. By providing interpretable topological metrics that directly correlate with pavement deterioration severity, this approach bridges the gap between advanced computer vision techniques and practical road maintenance decision-making. The proposed framework highlights the potential of automated topological analysis to enhance strategic infrastructure management by delivering actionable insights into crack patterns and their implications for structural health.
The functional relationship between the intensity of spatial use floor area ratio (FAR) and road network performance is fundamental in the context of rapid urban development. This study aims to quantify and model the spatial relationship between the actual FAR, deterioration of road performance—degree of saturation (DS) and side barriers (HS), on the main road corridor of Parepare City, using a deductive quantitative approach. The analysis begins with the collection of FAR and DS data referring to Indonesian road performance guidelines (PKJI, 2023), followed by the estimation of a Spatial lag regression model (SLM) to internalize the dimensions of spatial dependence. SLM model shows strong explanatory power ($R^2$ = 0.78), empirically confirming a positive and significant correlation between improvement FAR and DS. A key finding is the validation of a significant positive spatial autocorrelation ($\rho$ = +0.387, $p <$ 0.01), which proves that congestion is the result of spillover effects between connected segments, not an isolated phenomenon. These results justify that traffic interventions should be network-based. Furthermore, this study applies predicted scenarios of FAR increase (+10%, +20%, and maximum zoning limit 4.0). The results of this scenario are crucial, increasing FAR to the maximum zoning limit drastically predicts total functional failure in most segments (predicted to reach LOS E and F), especially in residential zones that show the highest FAR sensitivity ($\beta_{FARtotal}$ = +0.200). The main contribution of this study is to provide an adaptive model to determine FAR based on a critical performance threshold (DS$_{max}$ = 0.75). Policy implications recommend a holistic integration between spatial planning and transportation regulations, demanding an immediate revision of the maximum FAR limit (as mandated on the priority map) to a sustainable FAR, as well as the implementation of network-based mitigation strategies, rather than point-based, to manage the urban mobility crisis sustainably.
Conventional Intelligent Speed Assistance (ISA) systems primarily rely on static map data or camera-based sign recognition, which limits their adaptability to dynamic driving conditions and real-time speed regulation. To address this gap, this paper proposes an AIbased Intelligent Speed Limiting System (ISLS) that integrates GPS, digital map APIs, and real-time vehicular data to predict and adjust vehicle speed proactively. The system employs a Long Short-Term Memory (LSTM) model for predictive speed adaptation based on upcoming road geometry, traffic context, and environmental inputs. A reinforcement learning-based control module ensures smooth and safe throttle or braking actions according to predicted limits. The design further incorporates hardware-level safety through isolation circuits and protective elements verified by simulation. A Python-based experimental testbed validates the proposed method in terms of response time and speed deviation; the results show advantages of the proposed method over the classical ISA systems. Hence, the proposed ISLS advanced a step closer to an adaptive, context-aware, and safety-sensitive speed control of the vehicle.
As the volume of dry bulk and other cargo handles at ports rises, the question of guaranteeing sustainability of ports has developed into an essential subject globally. In Malaysia, green port initiatives were only introduced by the government in 2016. Therefore, limited research on green ports can be traced in the open literature. Compared to other cargo types, dry bulk cargo handling is regarded as one of the more demanding undertakings to the environment. Hence, the aim of this study is to ascertain the dry bulk terminals’ key green port performance determinants by using the Delphi method to identify the key determinants and Analytic Hierarchy Process (AHP) to quantitatively weigh the selected determinants, through a case study of Lumut Port. The main output of this study is a decision-making model that can be utilised to guide the efforts towards achieving a green port status by Lumut Port or other dry bulk terminals with similar characteristics. The research found that the cluster with the highest weightage is the Water Pollution Management Cluster (WPM), followed by the Environmental Awareness and Training cluster and the Air Pollution Management cluster. These three clusters presented 57.14% of the total weightage in achieving a green port status, which is worthy to be given serious consideration.
North Sulawesi’s geography includes both its mainland and the island regencies of Sangihe, Talaud, and Sitaro (Siau Tagulandang Biaro). This study examines the economic and service disparities between the island regencies of Sangihe, Talaud, and Sitaro and mainland North Sulawesi. In 2023, the Human Development Index (HDI) and Gini ratio for these island regencies were below the averages for both Indonesia and North Sulawesi. To address these gaps, the Ministry of Transportation implemented subsidized sea and ferry transport programs. Using a mixed-methods approach, the research combines a quantitative analysis of purchasing power with a qualitative review of relevant regulations. The study surveyed residents in the island regencies who utilize these subsidized services to assess their Ability to Pay (ATP) and Willingness to Pay (WTP). The findings reveal a significant gap: ATP for both passengers and freight is consistently lower than WTP, indicating a willingness to pay more than they can currently afford. Further analysis shows a disconnect between these ATP-WTP values and government-regulated fares, creating inconsistencies that influence consumer travel choices. This lack of alignment between bottom-up demand for affordable transport and top-down regulatory frameworks has led to inefficient service integration. While multimodal transshipment could offer a potential solution, its implementation is currently hindered by significant geographic and regulatory challenges, perpetuating the economic and service disparities faced by the island regencies.
Open Access
News 
Journal Metrics
Journal Overview
Associated Journals