Institute for Transportation

About the research

Road weather information systems (RWIS), in both stationary and mobile forms, have become increasingly popular in recent decades for their ability to collect and disseminate road weather and surface data. In addition to meteorological measurements, highway agencies rely heavily on RWIS imagery data to guide winter road maintenance (WRM) operations. However, the analysis of imagery data is still performed manually by trained personnel. Moreover, the limited number of stationary RWIS stations and the infrequent deployment of mobile RWIS units result in significant spatial gaps along the highway network. In our previous project, we developed methodologies based on convolutional neural networks (CNNs) to automatically recognize road surface conditions (RSC) from dash camera imagery and employed regression kriging (RK) to estimate RSC in unmonitored areas using limited point measurements. These methods demonstrated feasibility and robustness in real-world case studies. Building on these efforts, this project aimed to further advance CNN development specifically for stationary RWIS imagery and assess its reliability using explainable artificial intelligence (XAI) techniques, including SHapley Additive exPlanations (SHAP) and class activation map (CAM)-based methods. Additionally, to automatically estimate snow coverage ratios from stationary RWIS imagery, two distinct deep learning-based computer vision techniques, pix-to-pix generative adversarial network and semantic segmentation, were employed. Furthermore, the RK method was revisited to better accommodate a wider range of weather events while considering their variability, and the potential monetary benefits of this approach were also explored. To address the limitations of RK in handling categorical variables, a novel geostatistical method, namely nested indicator kriging (NIK), was developed to interpolate RSC in unmonitored areas directly using CNN classification results. These methods were evaluated using data from two major highways, Interstate 35 and Interstate 80 in Iowa, US, spanning a five-year period and encompassing over 20,000 images. The results demonstrated high accuracy and reliability. Additionally, a web application was developed to integrate these methods, offering real-time monitoring, estimation, and historical data archiving. This project equips decision-makers with a powerful tool to implement WRM activities more swiftly, efficiently, and cost-effectively, ultimately promoting a safer, more mobile, and sustainable winter transportation system.

About the research

Department of transportation (DOT) maintenance supervisors utilize a variety of tools including maintenance decision support systems (MDSS) to gain a better understanding of current and future road surface conditions (RSC) during winter weather. MDSS automatically attempt to deduce current RSC based on road weather information system (RWIS) and other data to a greater or lesser degree of accuracy. Although current MDSS implementations present highway camera imagery, they typically do not incorporate automated camera image recognition in order to improve the MDSS assessment of winter RSC. Thus, there can be discrepancies between the road weather conditions in camera images and MDSS RSC assessments. For example, an MDSS assessment may determine that a highway is clear whereas associated camera images show snow or vice versa. Such discrepancies can lead to a loss of confidence, system criticism, and noncompliance with MDSS recommendations. From that point of view, the integration of automated RSC camera image recognition into MDSS implementations can have a number of benefits:

• Better RSC assessment performance
• Better road treatment recommendations owing to better RSC identification
• Improved MDSS use and compliance with system recommendations owing to user confidence

Recent research in RSC identification has applied convolutional neural networks (CNN) and related techniques to the winter RSC identification problem. The National Center for Atmospheric Research (NCAR) is interested in transitioning these automated RSC identification techniques from the research community to the DOT community. Since NCAR has significant experience in the research-to-operations arena, the NCAR team worked with Aurora Program members to develop a set of recommendations for transitioning the relevant technology to MDSS applications.

Even though recent research efforts seem quite successful, the NCAR team was also interested in the potential to improve the initial CNN RSC identification by incorporating additional relevant data such as the following:

• RWIS precipitation/temperature data
• Vehicle speed/volume data

About the research

The objectives of this project were to determine the relationship between weather conditions and roadway friction measurements as observed in the laboratory, determine whether it is possible to standardize friction measurements coming from multiple friction sensors for identical weather conditions and roadway pavement types, determine whether the relationship between weather and roadway friction found in the laboratory is analogous to the relationship between weather and pavement friction found in practice on highways, and model roadway friction using weather conditions to predict it at sites where friction measurements may not be available.

The objectives were accomplished through cold laboratory testing of stationary friction sensors, standardizing friction measurements from multiple stationary and mobile friction sensors, using meteorological measurements from Colorado and Minnesota to infer road friction conditions, and conducting a friction wheel measurement analysis using data from Sweden.

Key findings from this effort include the following:

• Machine learning models can be created using data from friction sensors in the cold laboratory that exhibit a good mean absolute error in predicting the laboratory friction response to meteorological conditions set in the laboratory, but the models have a higher mean absolute error when applied to data from the field. For this reason, the researchers do not recommend using the model developed in the laboratory with field data.
• Collocated road weather information system (RWIS) and stationary friction sensor data can be used to develop state-specific friction models using machine learning techniques. These models can then be used to provide a synthetic friction estimate at RWIS sites that are not equipped with stationary friction sensors. The accuracy of the predictions can be determined at sites where friction sensors are available. The accuracy is improved when water thickness and/or snow thickness are available.
• RWIS measurements including air temperature, surface temperature, dew point temperature, relative humidity, and road condition measurements including road state, water thickness, and snow thickness can be used to derive an accurate friction model that targets observed friction values.
• Friction values from multiple sensor types are close in magnitude when friction is high, but when friction values drop, agreement among sensors is variable. To standardize the measurements from multiple friction sensors, friction values from multiple sensors can either be averaged or associated with a set of friction categories.

About the research

Connected vehicles (CVs) are equipped with advanced technologies that allow them to communicate with infrastructure, other vehicles, and external systems. Meanwhile, transportation agencies maintain networks of road weather information systems (RWIS) that monitor weather and pavement surface conditions and communicate these conditions to agency personnel in order to facilitate maintenance operations and decision-making. Integrating CV technology into RWIS infrastructure has the potential to significantly improve road safety by providing real-time weather information to drivers.

This project white paper presents the findings to date of the Integration of Connected Vehicle (CV) and Road Weather Information System (RWIS) Technologies project. The project reviewed the literature related to CV-RWIS integration, analyzed ongoing projects carried out by departments of transportation (DOTs) on CVs and RWIS, surveyed state and local transportation agencies regarding their investments in RWIS technologies to facilitate integration with CVs, conducted follow-up interviews with selected agencies, and prepared a summary of current challenges to CV-RWIS integration and recommendations for future research.