Arkansas Study Used Automatic Vehicle Identification System to Auto-Calibrate Weigh-In-Motion Stations.
Weigh-in-Motion (WIM) systems are effective in capturing weight and axle configurations of vehicles and can be served as essential inputs for evaluating transportation infrastructure performance. Auto-calibration of sensors, which is an algorithmic procedure by which weights measured by the WIM sensor are adjusted by calibration factors that are periodically calculated based on presumed traffic characteristics, is often performed to ensure weights are measured accurately. This project developed a new form of auto-calibration to measure the weights of the same truck, tracked by Automatic Vehicle Identification (AVI) across multiple WIM sites, to generate a reference weight and calibration factor. Two data collection efforts were undertaken in March of 2018 and 2019 to gather WIM, static scale, and video data for algorithm validation resulting in a sample of approximately 500 trucks to use for algorithm performance evaluation.
The proposed AVI-based auto-calibration method consisted of first, matching AVI-tracked trucks to WIM Per Vehicle Records (PVR), and second, applying a calibration procedure that measures the weights of the same truck using AVI across multiple WIM sites to produce a reference weight and calibration factor. The proposed AVI-based approach in this project was compared to the Arkansas Department of Transportation (ARDOT) and Minnesota Department of Transportation (MnDOT) approaches for a set of six Arkansas WIM sites, where ARDOT generated calibration factors based on the measured front axle weights (FAW) averaged for a sample of 50 five-axle tractor trailers and MnDOT expanded on that approach by defining three FAW references based on Gross Vehicle Weight (GVW) bins and applying correction factors when sample sizes are small. The researchers collected WIM PVR at each WIM site, AVI data from a national truck GPS data provider, and static weight recordings at Arkansas Highway Police weight enforcement sites during two data collections. In total, approximately 500 truck matches from WIM to static scale locations were identified, which were used to evaluate performance of the three auto-calibration approaches.
- Calibration accuracy: Without any form of auto-calibration, FAW errors ranging from 24 to 85 percent were observed, with GVW error in the same range. The ARDOT method reduced errors to between 12 and 16 percent for FAW and 14 to 29 percent for GVW. The MnDOT method reduced errors to between 11 and 26 percent for FAW and 11 to 41 percent for GVW. The AVI-based method reduced errors to between 10 and 35 percent for FAW and 16 and 35 percent for GVW.
- Auto-calibration cost savings: If AVI data were to be used in-place of on-site calibration methods, 98 percent cost savings could be realized (Table 1). Since it may not be realistic to assume that all WIM sites can be calibrated using AVI data, a combination of on-site (field data collection) and AVI-based methods is more realistic, which translates into an anticipated 30 percent cost savings compared to on-site only calibration.
- Return on Investment (ROI): The ROI for this project was estimated as 63 percent cost savings for ARDOT if auto-calibration using AVI were to replace on-site calibration requirements that involve test trucks of known weight or use of static scales for weight comparisons.
Table 1. Summary of Estimated Calibration Costs by Calibration Method (50 WIMs)
|Annual Cost (Labor cost excluded)
|Test Trucks (status-quo method)
|Auto-Calibration Using AVI-Based Data Collection
|Auto-Calibration Using Field and AVI-Based Data Collection (30% using AVI and 70% using test trucks)