The field of traffic monitoring involves program planning, design and evaluation, and data collection, analysis, and reporting. This session presents research aimed at improving traffic monitoring practices for all modes, with a focus on emerging traffic monitoring data acquisition and processing methods and tools.
Development of Statewide AADT Estimation Model from Short-Term Counts: A Comparative Study for South Carolina
Sakib Khan, Clemson UniversityShow Abstract
Sababa Islam, Clemson University
MD Zadid Khan, Clemson University
Kakan Dey, West Virginia University
Mashrur Chowdhury, Clemson University
Nathan Huynh, University of South Carolina
Annual Average Daily Traffic (AADT) is one of the most important parameters used in traffic engineering analysis. Department of Transportations (DOTs) continually collect traffic count using both permanent count stations (i.e., Automatic Traffic Recorders or ATR) and temporary short-term count stations. In South Carolina, 87% of the ATRs are located on interstates and arterial highways. For the majority of the secondary highways (i.e., collectors and local roads), AADT is estimated based on short-term counts. This paper develops AADT estimation models for different roadway functional classes with two machine learning techniques: Artificial Neural Network (ANN) and Support Vector Regression (SVR). The models aim to predict AADT from short-term counts. The results are compared at first, with each other, to identify the best model. Then, the results of the best model are compared with the traditional factor based method and regression method. The comparison reveals the superiority of SVR for AADT estimation for different roadway functional classes over all other methods. Among all developed models for different functional roadway classes, SVR-based model showed minimum Root Mean Square Error (RMSE) and Mean Absolute Percentage Error for the interstate/expressway functional class. This model also showed higher R-squared value compared to traditional factor model and regression model. The SVR models are validated for each roadway functional class using the 2016 ATR data, and selected short-term count data. The validation result proves that the SVR-based AADT estimation models can be used by SCDOT to reliably predict AADT from the short-term counts.
Comparison of Two Short-Period Traffic Count Duration/Cycle Specifications in the Accuracy of Their Predictions of Annual Average Daily Traffic at Coverage Stations
Michelle Edwards, Tennessee Department of TransportationShow Abstract
Daniel Badoe, Tennessee Technological University
David Lee, Tennessee Department of Transportation
Transportation performance data among others guide the allocation of funds to states for investments in transportation facilities and services. A key performance measure in this regard is the annual average daily traffic (AADT) which is estimated from daily volume counts made over a year. Thus, U.S. Department of Transportation requires that each U.S. state obtain estimates of AADT for the different sections of their monitored road network. Each state department of transportation therefore has a traffic monitoring program for the collection of volume data. Sub-elements of this program include a permanent traffic count (PTC) program and a short period traffic count (SPTC) program. Traffic volumes collected in the SPTC program are adjusted to AADT estimates using seasonal factors estimated from data collected at PTC stations. The typical duration of SPTCs is 48 hours undertaken on a three-year cycle. Tennessee undertakes 24-hour counts on a one-year cycle, which translates into a much greater workload for count field-staff raising the question of whether there is an accuracy benefit to Tennessee’s count protocol that may help discount its onerousness . Thus, the objective of this research was to compare the accuracy of AADT estimates obtained from the above two alternative SPTC duration/cycle specifications. The PTC volumes used in the research were collected by Tennessee Department of Transportation (TDOT) in Years 2013 to 2015. The study results led to the conclusion that AADT-estimates based on 48-hour counts on a three-year cycle are more accurate than those given by 24-hour counts undertaken yearly.
Viability Evaluation of Estimating Axle Factors and Axle Classes from Vehicle Length Data
Raul Avelar, Texas A&M Transportation InstituteShow Abstract
Tomas Lindheimer, Texas A&M Transportation Institute
Sruthi Ashraf, Texas A&M Transportation Institute
This study developed methods to estimate axle factors and vehicle class from length-based data streams. A set of eight methods was proposed and evaluated in different testing schemes intended to observe performance on homogeneous and heterogeneous data.
The initial analysis used length-based data from 61 sites in Wisconsin. The research team compared performance of the methods estimating axle factors and vehicle class proportions. Performance was comparable and consistent between homogeneous and heterogeneous subsets of data.
The research team selected two methods for a final round of analysis due to their accuracy and robustness to heterogeneity. For the final round of analysis, the research team assembled a multistate dataset using data from Wisconsin and from 14 other states represented in a dataset from the Long Term Pavement Performance (LTPP) program. The final round of analysis compared performance under different seasons, facility type, and road character (urban vs. rural). Performance of the two identified methods was deemed appropriate and they are recommended for implementation.
Video Tool for Manually Extracting Complex Traffic Data
Abhilasha Saroj, Georgia Institute of Technology (Georgia Tech)Show Abstract
Nishu Choudhary, Georgia Institute of Technology (Georgia Tech)
Han Gyol Kim, Georgia Institute of Technology (Georgia Tech)
Angshuman Guin, Georgia Institute of Technology (Georgia Tech)
Michael Rodgers, Georgia Institute of Technology (Georgia Tech)
Michael Hunter, Georgia Institute of Technology (Georgia Tech)
Driver behavior studies often require the analysis of highly specific and customized observational data attributes. Due to the complex nature of behavioral data collection, typically requiring some degree of customization depending on the study, there is a lack of generic tools available to transportation professionals to assist in the data collection process. This paper presents a python-based software application “GT-MVP” designed to provide a user-friendly interface to collect complex video-based traffic data. GT-MVP’s graphical user interface allows users to play multiple videos and operate them synchronously using common controls, to review the extracted data in real-time, and to correct errors easily. GT-MVP has been used to collect data to study vehicle’s blocking behavior at intersections. Compared to previous approaches used to collect behavioral data required for this study, GT-MVP took 65 % lesser time and reduced the missed detection rate. GT-MVP interface can be modified to collect complex traffic data for different traffic studies and can also be used to improve efficiency of collection of basic traffic data such as vehicle counts and is available to the community as an open source software.
Challenges in Monitoring Regional Trail Traffic
Greg Lindsey, University of Minnesota, Twin CitiesShow Abstract
Lila Singer-Berk, University of Minnesota, Twin Cities
Jeffrey Wilson, IUPUI
Eric Oberg, Rails-To-Trails Conservancy
Tracy Loh, George Washington University
This case study reports results of traffic monitoring at 30 locations on a 972-mile shared-use trail network across the east-central United States. We illustrate challenges in adapting the principles in the Federal Highway Administration’s Traffic Monitoring Guide to a regional trail network. We make four contributions: (1) we use factor analysis and k-means clustering to implement a stratified random process for selecting monitoring sites; (2) we illustrate quality assurance procedures and the challenges of obtaining valid results from a multi-state monitoring system; (3) we describe variation in trail traffic volumes across five land use classes in response to daily weather and seasons and present monthly adjustment factors for use in extrapolation of counts; and (4) we report two performance measures for the network: annual average daily trail traffic and trail miles traveled. The Rails to Trails Conservancy, in collaboration with the Industrial Heartland Trails Coalition, deployed thirty automated passive infrared traffic monitors late in 2015 through early 2017. Site-specific regression models were used to estimate and impute missing daily traffic volumes. The effects of weather generally were consistent across land use classes but the effects of temporal variables, including weekend and season-of-year, varied. Results confirm both the applicability of the FHWA principles and the difficulties associated with implementing them in a regional context.