This research addresses operational disconnects and knowledge gaps related to traffic data collection in rural areas by developing a low-cost 3D-printed and open-source traffic counter (CTRA). Conventional pneumatic tube-based systems, which are still in use by transportation agencies across the United States because of their affordability, simply do not work on gravel roads and have difficulty counting non-motorized users and differentiating non-traditional vehicles from conventional motor vehicles. CTRA was developed and field tested at the University of Alaska Fairbanks and designed to provide a video-based data collection system that overcomes the limitations of other traffic counting devices. A count rate of 100% was achieved during the calibration process. Other than electronic hardware, most pieces of hardware can be printed on a 3D printer to form a simple and robust case and mounting system and only straps are needed to secure the counter to a fixed object. Because of its relatively simple and affordable design, CTRA could also be used for STEM and educational activities in schools and other related programs.

In this study, we focus on automating road marking extraction from the HDOT MLS point cloud database, managed by Mandli. Mandli is a company specializing in highway data collection, including LiDAR. Mandli has cooperated with various Department of Transportation throughout the United States. Here, we focus on infrastructure elements related to non-motorized travel modes, supporting the ongoing Complete Streets efforts in Hawaii. Point cloud data include different colors that represent differences in elevation and intensity values. Based on a visual inspection, road markings can be observed within these point clouds. The long-term objective of this study is to develop a framework and approach for automating the detection of these infrastructure elements, based on deep learning approaches. For this project, a YOLOv5 (You Only Look Once version 5) image object detection model was trained with the HDOT point cloud data. YOLO is a family of deep learning models designed for fast object detection; the latest published version is the 5th version. The focus here is on non-motorized objects, such as crosswalks, bike lanes and bike boxes. The same approach can be extended to other markings as well, which we plan for subsequent studies.

A numerical tool was developed that helps forecast when thaw occurs at depths in a road embankment during spring thaw in regions that experience seasonal freeze and thaw. The tool is a Excel spreadsheet that uses a single adjustable parameter, and is driven by time series air temperature data. The model agrees well with archived data of subsurface temperatures at five different highway locations in Alaska.

This report presents the Mobile Unit for Sensing Traffic (MUST) project, which addresses the critical need for comprehensive traffic data in tribal and rural areas that often suffer from disproportionately high rates of traffic incidents due to limited infrastructure and resources. By implementing a low-cost, user-friendly data collection system, the MUST project aimed to provide real-time traffic information to significantly enhance traffic safety. The pilot installation in Yakama Nation, Washington, at a high-traffic intersection known for frequent accidents, marks a crucial step in this initiative. Collaboration with the Yakama Nation's Tribal Traffic Safety Coordinator and Yakama Power ensured the successful installation and operation of the sensor on a strategically selected telephone pole. Equipped with advanced machine learning technology, the MUST system collects detailed data on traffic flow, road surface conditions, and environmental factors such as temperature and humidity, visualized through a sophisticated dashboard for real-time monitoring and data-driven decision-making. This system allows for the identification of high-risk areas and the implementation of targeted safety measures, such as improved signage and road maintenance, while addressing specific concerns like pedestrian safety, visibility issues due to heavy fog, and speeding. By providing a robust dataset previously unavailable, the MUST project supports the Yakama Nation’s efforts to understand and mitigate traffic safety issues, ultimately enhancing the overall safety and quality of life for the community. This pilot project serves as a model for other tribal and rural areas looking to leverage advanced technology to improve their transportation safety infrastructure.

The report presents a unique contrast for several Machine Learning approaches aiming at understanding pedestrian fatal collisions. Four classification techniques are applied to assess how roadway features mainly correlate to pedestrian fatal crashes: Logistic Regression, Nearest Neighbor Classification, Decision Tree, and Random Forest Classifier. The data used in this project was collected from the Highway Safety Information System (HSIS) database, which provides both collision data for the entire state of Washington and roadway characteristics for all state highways. Each of the four modeling approaches was implemented using K-fold cross-validation, a process that allows choosing the best parameters for the model. Their results were evaluated and then compared in terms of accuracy score and confusion matrices for the testing data set. It was found that the Decision tree had consistent results and the best performance among all models, showing how the distinct predictors relate to each other to predict fatal pedestrian collisions.

Recent technological advancements in computer vision algorithms and data acquisition devices have greatly facilitated research activities towards enhancing traffic sensing for traffic safety performance improvements. Significant research efforts have been devoted to developing and deploying more effective technologies to detect, sense, and monitor traffic dynamics and rapidly identify crashes in in Rural, Isolated, Tribal, or Indigenous (RITI) communities. As a new modality for 3D scene perception, Light Detection and Ranging (LiDAR) data have gained increasing popularity for traffic perception, due to its advantages over conventional RGB data, such as being insensitive to varying lighting conditions. In the past decade, researchers and professionals have extensively adopted LiDAR data to promote traffic perception for transportation research and applications. Nevertheless, a series of challenges and research gaps are yet to be fully addressed in LiDAR-based transportation research, such as the disturbance of adverse weather conditions, lack of roadside LiDAR data for deep learning analysis, and roadside LiDAR-based vehicle trajectory prediction. In this technical report, we focus on addressing these research gaps and proposing a series of methodologies to optimize deep learning-based feature recognition for roadside LiDAR-based traffic object recognition tasks. The proposed methodologies will help transportation agencies monitor traffic flow, identify crashes, and develop timely countermeasures with improved accuracy, efficiency, and robustness, and thus enhance traffic safety in RITI communities in the States of Alaska, Washington, Idaho, and Hawaii.

Ice roads and bridges across rivers, estuaries, and lakes are common transportation routes during winter in regions of the circumpolar north. Ice thickness, hydraulic hazards, climate variability and associated warmer air temperatures have always raised safety concerns and uncertainty among those who travel floating ice road routes. One way to address safety concerns is to monitor ice conditions throughout the season. We tested ground penetrating radar (GPR) for its ability and accuracy in measuring floating ice thickness under three specific conditions: 1) presence of snow cover and overflow, 2) presence of snow cover, and 3) bare ice, all common to Interior Alaska rivers. In addition, frazil ice was evaluated for its ability to interfere with the GPR measurement of ice thickness. We collected manual ice measurements and GPR cross-sectional transects over 2 years on the Tanana River near Fairbanks, Alaska, and for 1 year on the Yukon River near Tanana, Alaska. Ground truth measurements were compared with ice thickness calculated from an average velocity model created using GPR data. The error was as low as 2.3–6.4% on the Yukon River (Condition 3) and 4.6–9.5% on the Tanana River (Conditions 1 and 2), with the highest errors caused by overflow conditions. We determined that certain environmental conditions such as snow cover and overflow change the validity of an average velocity model for ice thickness identification using GPR, while frazil ice accumulation does not have a detectable effect on the strength of radar reflection at the ice-water interface with the frequencies tested. Ground penetrating radar is a powerful tool for measuring river ice thickness, yet further research is needed to advance the ability of rural communities to monitor ice thickness using fewer time-intensive manual measurements to determine the safety of ice cover on transportation routes.

This study examined school travel safety and risk and explored the potential differences between conditions that are present today with those that existed nearly two decades ago, when the Transportation Research Board published its landmark study on school travel safety. For this study, thirty transportation professionals were interviewed and a twenty-year crash data set from the Fatality Analysis Reporting System (FARS) was analyzed. The responses from the interviews were separated into ten common themes. The three most mentioned themes were education programs, concerns of roadway environments, and school bus safety. Based on the responses, concerns about the roadway environment, poor driver behavior, and the role of parents on mode choice have not changed in the last twenty years; however, safety education programs, vehicle centric travel, community planning, and pick up/drop off safety have evolved over time. With regard to the FARS data set, which was used as a benchmark to assess school transportation safety, the overall trends indicate that the trip to and from school remains a relatively safe activity, particularly along rural facilities where positive results were identified across four key metrics. Along urban facilities, slightly increasing trends were observed in the annual number of fatalities and in the number of non-motorists involved in a fatal crash, suggesting that opportunities remain to enhance and to improve the travel environment for school children.

This report documents the research activities to develop a drone-based computer vision-enabled vehicle dynamic safety performance monitoring in Rural, Isolated, Tribal, or Indigenous (RITI) communities. The acquisition of traffic system information, especially the vehicle speed and trajectory information, is of great significance to the study of the characteristics and management of the traffic system in RITI communities. The traditional method of relying on video analysis to obtain vehicle number and trajectory information has its application scenarios, but the common video source is often a camera fixed on a roadside device. In the videos obtained in this way, vehicles are likely to occlude each other, which seriously affects the accuracy of vehicle detection and the estimation of speed. Although there are methods to obtain high-view road video by means of aircraft and satellites, the corresponding cost will be high. Therefore, considering that drones can obtain high-definition video at a higher viewing angle, and the cost is relatively low, we decided to use drones to obtain road videos to complete vehicle detection. In order to overcome the shortcomings of traditional object detection methods when facing a large number of targets and complex scenes of RITI communities, our proposed method uses convolutional neural network (CNN) technology. We modified the YOLO v3 network structure and used a vehicle data set captured by drones for transfer learning, and finally trained a network that can detect and classify vehicles in videos captured by drones. A self-calibrated road boundary extraction method based on image sequences was used to extract road boundaries and filter vehicles to improve the detection accuracy of cars on the road. Using the results of neural network detection as input, we use video-based object tracking to complete the extraction of vehicle trajectory information for traffic safety improvements. Finally, the number of vehicles, speed and trajectory information of vehicles were calculated, and the average speed and density of the traffic flow were estimated on this basis. By analyzing the acquiesced data, we can estimate the traffic condition of the monitored area to predict possible crashes on the highways.

Travel during winter months remains particularly problematic in the Pacific Northwest due to the regular occurrence of inclement weather in the form of snow and ice during freezing and sub-freezing conditions. For travelers and commuters alike, vehicle traction in the form of studded tires serves to provide an added level of driving confidence when weather conditions deteriorate. However, recurring studded tire usage causes damage to the roadway infrastructure in the form of surface wear and rutting over time. Left unattended, this damage contributes to challenging and potentially dangerous driving conditions in the form of standing water and the increased potential for hydroplaning. Currently, an efficient and automated method to collect site-specific studded tire traffic volumes is lacking. While studded tire usage can be locally estimated based on manual roadway traffic counts, parking lot counts, or household surveys, the lack of real-world traffic volumes prevents the fine-tuning of roadway deterioration models that measure performance and estimate infrastructure life. This project tested the use of off-the-shelf sound meters to determine if an acoustic method could be developed to measure studded tire volumes. Based on the results, a prediction model was developed to allow for data-driven solutions that will benefit local transportation officials, planners, and engineers responsible for managing highways and roadways.