Research Team: Francois Dion (lead), Mingyuan Yang, and Anthony Patire
UC Campus(es): UC Berkeley
Problem Statement: Determining where trucks are traveling is crucial for the planning and maintenance of transportation networks. In California, information regarding truck movements is primarily derived from a network of fixed traffic monitoring stations. While data collected from these stations can be used to classify passing trucks, determine their travel direction, and assess their proportion within the general traffic, they provide limited information about trip origins and destinations and the routes taken between stations. Estimating truck movements within a region thus largely depends on extrapolating data between known collection points. Although this can be done with relative ease in simple networks offering few alternate routes, it can be a difficult task in complex networks without a multitude of supporting observation points. A potential solution to the above problem is using vehicle tracking data, also known as probe data, supplied by third-party vendors to fill in gaps in truck monitoring. This data is collected from individual onboard vehicle monitors or GPS-enabled navigation devices in the vehicle. It is typically used by fleet operators to manage their business, but it can also be used to provide accurate information about truck movements not available from roadside monitors.
Project Description: This report describes how current methods of estimating truck traffic volumes from existing fixed roadway sensors could be improved by using tracking data collected from commercial truck fleets and other connected technology sources (e.g., onboard GPS-enabled navigation systems and smartphones supplied by third-party vendors). Using Caltrans District 1 in Northern California as an example, the study first reviews existing fixed-location data collection capabilities and highlights gaps in the ability to monitor truck movements. It then reviews emerging data sources and analyzes the analytical capabilities of StreetLight 2021, a commercial software package. The study then looks at the Sample Trip Count and uncalibrated Index values obtained from three weigh-in-motion (WIM) and twelve Traffic Census stations operated by Caltrans in District 1. The study suggests improvements to StreetLight’s “single-factor” calibration process which limits its ability to convert raw truck count data into accurate traffic volume estimates across an area, and suggests how improved truck-related calibration data can be extracted from the truck classification counts obtained from Caltrans’ WIM and Traffic Census stations. The report compares uncalibrated StreetLight Index values to observed truck counts to assess data quality and evaluates the impacts of considering alternate calibration data sets and analysis periods. Two test cases are presented to highlight issues with the single-factor calibration process. The report concludes that probe data analytical platforms such as StreetLight can be used to obtain rough estimates of truck volumes on roadway segments or to analyze routing patterns. The results further indicate that the accuracy of volume estimates depends heavily on the availability of sufficiently large samples of tracking data and stable and representative month-by-month calibration data across multiple reference locations.