OHS: Estimating Traffic Volume Using Ubiquitous Probe Vehicle Data (AADT) M3DA-2023-00-00-06

Abstract OHS, Estimating Traffic Volume Using Ubiquitous Probe Vehicle Data 10/01/22 – 09/30/23 The project aims to fill the gaps in traffic volume data on Kentucky’s roadway network by employing proven machine learning method and ubiquitous probe vehicle data. The project will improve the accuracy and completeness of roadway database and help the KYTC become compliant to the HSIP Final Rule requirement. Annual Average Daily Traffic (AADT) is fundamental to a wide range of transportation applications. The recent Highway Safety Improvement Program (HSIP) Final Rule requires states to have access to AADT data for all public paved roadways, including local roads by 2026. KYTC’s Traffic Count Reporting System plays a central role in collecting and providing this information through continuous count stations, short-term traffic counters, and classification data collection stations at selected locations on the state-maintained highways. However, AADT on many roads, especially local roads, is often unavailable due to limited resources. A methodology is needed to provide accurate estimates of traffic volumes on those unmonitored segments. Building on the findings of a previous KTC study on local roads ADT estimation, this study will employ more advanced machine learning methods and improved probe vehicle data to produce more accurate volume estimates. SECTION C: Goals and Objectives: 1. To estimate statewide AADT on all roads, especially local roads, in Kentucky. 2. To evaluate the accuracy of the estimates using continuous count data. 3. The study would assist KYTC’s effort in meeting HSIP requirement and providing input to a wide range of transportation applications. SECTION D: Strategies and Activities: We propose to adapt a machine-learning method to estimate AADT on Kentucky highways, including local roads. To implement the model, various data sets, such as roadway geometry, probe speeds and counts, vehicle registrations, as well as demographic and lane use data will be collected and processed. The variables with significant effect on AADT will be selected and the model will be calibrated to achieve the optimal performance. Traffic counts obtained from the permanent counters will be used to evaluate the model performance.