Denver Tolliver

Accident prediction model for public highway-rail grade crossings

Considerable research has focused on roadway accident frequency analysis, but relatively little research has examined safety evaluation at highway-rail grade crossings. Highway-rail grade crossings are critical spatial locations of utmost importance for transportation safety because traffic crashes at highway-rail grade crossings are often catastrophic with serious consequences. The Poisson regression model has been employed to analyze vehicle accident frequency as a good starting point for many years. The most commonly applied variations of Poisson including negative binomial, and zero-inflated Poisson. These models are used to deal with common crash data issues such as over-dispersion (sample variance is larger than the sample mean) and preponderance of zeros (low sample mean and small sample size). On rare occasions traffic crash data have been shown to be under-dispersed (sample variance is smaller than the sample mean) and traditional distributions such as Poisson or negative binomial cannot handle under-dispersion well. The objective of this study is to investigate and compare various alternate highway-rail grade crossing accident frequency models that can handle the under-dispersion issue. The contributions of the paper are two-fold: (1) application of probability models to deal with under-dispersion issues and (2) obtain insights regarding to vehicle crashes at public highway-rail grade crossings.

Analyzing Effects of Spring Highway Load Restrictions on North Dakota’s Agricultural Freight Flows

This paper describes a statewide agricultural freight transportation model that is used to estimate the benefits of improving North Dakotas state highway system by removing spring load restrictions. The transportation model measures changes in freight flows caused by truck weight regulations during the spring thaw cycle. The primary focus is on grain transportation. A geographic information system (GIS) network of federal, state, and county roads is developed to represent flows from fields to elevators and final destinations. The state is divided into 182 production zones on the basis of agricultural land use patterns. The data for the trip generation component of the model are derived from satellite imagery of crop layers in the state, with the use of GIS spatial analysis and algorithms developed for this purpose. The annual demand at elevators is estimated from grain movement reports filed with the State of North Dakota. Agricultural freight movement is modeled separately as two flows: internal-to-internal (i.e., from fields to elevators) and internal-to-external (i.e., from elevators to final destinations). CUBE transportation planning software is used to model the flows. In addition, the onion model concept used in Iowa is applied as a demand planning tool to capture the effects of spring load restrictions, which are dynamic and move from the southern to the northern part of the state during the spring thaw cycle. The costs of spring load limits are quantified as the increased distance and travel time caused by circuitous truck movements or as the reduced payload per trip.

Analyzing Investments Needed to Support Oil and Gas Production and Distribution

The purpose of this study was to forecast road investment needs in the oil- and gas-producing counties of North Dakota over the next 20 years in light of its expected population growth and the growth of oil and gas production. With the essential objective of quantifying the additional investments necessary for efficient year-round transportation of oil while providing travelers with acceptable roadway service, the study focused on roads owned or maintained by local governments, e.g., counties and townships. Impacts and funding needs were analyzed for three types of roads: paved, gravel, and graded and drained. The analysis was based on three main data sources: oil production forecasts, traffic data, and county road surveys. The forecast output of wells was routed over the road network to pipelines, with a detailed geographic information system model in which oil movements were represented as equivalent tractor–semitrailer trips that follow least-cost paths. The projected inputs of sand and water and the outbound movements of salt water to disposal sites were similarly routed. These predicted inbound and outbound movements were accumulated for each impacted segment. Movements of specialized equipment were included in the analysis. Several types of potential road improvements, including reconstruction and structural overlays, were analyzed in this study. The model developed in this research can be transferred to other states where new oil fields are opening and analysis is required for additional investment in highway infrastructure.

Decision Tree Approach to Accident Prediction for Highway–Rail Grade Crossings: Empirical Analysis

Highway–rail grade crossings (HRGCs) are critical spatial locations that are of utmost importance for transportation safety because traffic crashes at these locations are often catastrophic. Compared with traditional regression models, the decision tree is more advanced in its ability to handle large data sets, deal with missing values, and not require predefined underlying relationships between target variables and predictors. Thus the decision tree approach is explored in this study, which evaluates HRGC crashes. Because crashes at HRGCs are rare, the majority of data will have a zero-crash classification. A traditional decision tree method will have a bias toward the majority classification, which will result in a good prediction for the majority class but a relatively poor prediction for rare events. To improve model accuracy with the decision tree model, especially for forecasting rare events, previous probability and decision profit values are adjusted. Historical crash data of North Dakota State from 1996 to 2014 are analyzed, and factors relevant to HRGC crashes are investigated. Results reveal that 23 variables are considered as contributors to crashes at highway–rail crossings. Results indicate that railway traffic, highway traffic, and train speed are main influential factors and have a positive impact on crash likelihood. Presence of train detecting and advance warning systems is helpful in reducing crash risk. Predicting accuracies of the model are 84.1% and 77.2% for event class and nonevent class, respectively.

Hyperspectral imaging utility for transportation systems

The global transportation system is massive, open, and dynamic. Existing performance and condition assessments of the complex interacting networks of roadways, bridges, railroads, pipelines, waterways, airways, and intermodal ports are expensive. Hyperspectral imaging is an emerging remote sensing technique for the non-destructive evaluation of multimodal transportation infrastructure. Unlike panchromatic, color, and infrared imaging, each layer of a hyperspectral image pixel records reflectance intensity from one of dozens or hundreds of relatively narrow wavelength bands that span a broad range of the electromagnetic spectrum. Hence, every pixel of a hyperspectral scene provides a unique spectral signature that offers new opportunities for informed decision-making in transportation systems development, operations, and maintenance. Spaceborne systems capture images of vast areas in a short period but provide lower spatial resolution than airborne systems. Practitioners use manned aircraft to achieve higher spatial and spectral resolution, but at the price of custom missions and narrow focus. The rapid size and cost reduction of unmanned aircraft systems promise a third alternative that offers hybrid benefits at affordable prices by conducting multiple parallel missions. This research formulates a theoretical framework for a pushbroom type of hyperspectral imaging system on each type of data acquisition platform. The study then applies the framework to assess the relative potential utility of hyperspectral imaging for previously proposed remote sensing applications in transportation. The authors also introduce and suggest new potential applications of hyperspectral imaging in transportation asset management, network performance evaluation, and risk assessments to enable effective and objective decision- and policy-making.

Error sensitivity of the connected vehicle approach to pavement performance evaluations

Abstract The international roughness index (IRI) is the prevalent indicator used to assess and forecast road maintenance needs. The fixed parameters of its simulation model provide the advantage of requiring relatively few traversals to produce a consistent index. However, the static parameters also cause the model to under-represent roughness that riders experience from profile wavelengths outside of the model’s response range. A connected vehicle method that uses a similar but different index to characterise roughness can do so by accounting for all vibration wavelengths that the actual vehicles experience. This study characterises and compares the precision of each method. The field studies indicate that within seven traversals, the connected vehicle approach could achieve the same level of precision as the procedure used to produce the IRI. For a given vehicle and segment lengths longer than 50 m, the margin-of-error diminished below 1.5% after 50 traversals, and continued to improve further as the traversal volume grew. Practitioners developing new tools to evaluate pavement performance will benefit from this study by understanding the precision trade-off to recommend the best practices in utilising the connected vehicle method.

Accuracy enhancement of roadway anomaly localization using connected vehicles

Abstract The timely identification and localisation of roadway anomalies that pose hazards to the traveling public is currently a critical but very expensive task. Hence, transportation agencies are evaluating emerging alternatives that use connected vehicles to lower the cost dramatically and to increase simultaneously both the monitoring frequency and the network coverage. Connected vehicle methods use conventional GPS receivers to tag the inertial data stream with geospatial position estimates. In addition to the anticipated GPS trilateration errors, numerous other factors reduce the accuracy of anomaly localisation. However, practitioners currently lack information about their characteristics and significance. This study developed error models to characterise the factors in position biases so that practitioners can estimate and remove them. The field studies revealed the typical and relative contributions of each factor, and validated the models by demonstrating agreement of their statistics with the anticipated norms. The results revealed a surprising potential for tagging errors from embedded systems latencies to exceed the typical GPS errors and become dominant at highway speeds.

Wavelength sensitivity of roughness measurements using connected vehicles

ABSTRACT Researchers previously demonstrated that a roughness index called the road impact factor (RIF) is directly proportional to the international roughness index (IRI) when measured under identical conditions. A RIF-transform converts inertial signals from connected vehicle accelerometers and speed sensors to produce RIF-indices in real time. This research examines the relative sensitivities of the RIF and the IRI to variations in dominant profile wavelengths. The findings are that both indices characterise roughness from spatial wavelengths up to 2 m with equal sensitivity. However, the RIF-transform maintains its sensitivity when characterising roughness from wavelengths beyond that. The case studies used a certified inertial profiler to collect both RIF and IRI data simultaneously from five different pavement surface types. The RIF/IRI proportionality factors distributed normally among the profiles tested. This result affirms that the RIF and IRI generally agrees. However, differences in the dominant profile wavelength among pavements will produce some spread in the degree of roughness that the indices express.

Resolution Agile Remote Sensing for Detection of Hazardous Material Spills

Traffic carrying flammable, corrosive, poisonous, and radioactive materials continues to increase in proportion with the growth in their production and consumption. The sustained risk of accidental releases of such hazardous materials poses serious threats to public safety. Early detection of spills will potentially save lives, protect the environment, and thwart the need for expensive cleanup campaigns. Ground patrols and terrestrial sensing equipment cannot scale cost-effectively to cover the entire transportation network. Remote sensing with existing airborne and spaceborne platforms has the capacity to monitor vast areas regularly but often lacks the spatial resolution necessary for high accuracy detections. The emergence of unmanned aircraft systems with lightweight hyperspectral image sensors enables a resolution agile approach that can adapt both spatial and spectral resolutions in real time. Equipment operators can exploit such a capability to enhance the resolution of potential target materials detected within a larger field-of-view to verify their identification or to perform further inspections. However, the complexity of algorithms available to classify hyperspectral scenes limits the potential for real-time target detection to support rapid decision making. This research introduces and benchmarks the performance of a low-complexity method of hyperspectral image classification. The hybrid supervised–unsupervised technique approaches the performance of prevailing methods that are at least 30-fold more computationally complex.

Use of Connected Vehicles to Characterize Ride Quality

The United States relies on the performance of more than 4 million miles of roadways to sustain its economic growth and to support the dynamic mobility needs of its growing population. The funding gap to build and maintain roadways is ever widening. Hence the continuous deterioration of roads from weathering and usage poses significant challenges. Transportation agencies measure ride quality as the primary indicator of roadway performance. The international roughness index is the prevalent measure of ride quality that agencies use to assess and forecast maintenance needs. Most jurisdictions use a laser-based inertial profiler to produce the index. However, technical, practical, and budget constraints preclude that use for some facility types, particularly local and unpaved roads that make up more than 90% of the road network in the United States. This study expands on previous work that developed a method to transform sensor data from many connected vehicles to characterize ride quality continuously, for all facility types and at any speed. The case studies used a certified and calibrated inertial profiler to produce the international roughness index. A smartphone aboard the inertial profiler produced simultaneously the roughness index of the connected vehicle method. The results validate the direct proportionality relationship between the inertial profiler and connected vehicle methods within a margin of error that diminished below 5% and 2% after 30 and 80 traversal samples, respectively.