Case Study: CATT Lab & FHWA NPMRDS Conflation

The Federal Highway Administration's (FHWA) National Performance Management Research Data Set (NPMRDS) is a comprehensive database that provides probe-vehicle-based speed and travel time data on the National Highway System (NHS). This data is crucial for transportation agencies to monitor and manage the performance of roadways, ensuring efficient and reliable travel. The NPMRDS supports various performance measures, including travel time reliability, congestion, and freight movement, helping agencies meet federal regulations and improve overall transportation system performance. The Center for Advanced Transportation Technology Laboratory (CATT Lab) at the University of Maryland is the prime contractor responsible for delivering this essential product.

Conflating data at this scale was previously performed by college students, making it extremely manually intensive and leading to inconsistent results due to subjective interpretations. Automating this process introduces additional challenges, including:

• Different Network Coverage: The INRIX TMCs are more expansive than just the NHS network. Logic needs to determine when a TMC is part of the NHS versus a local road.
• Different Linework: The scale and precision of the digitized road centerline network from two different sources are never aligned due to the varying scales and precision at which they are collected. This requires fuzzy buffer distance tolerances to account for the offset networks.
• Different Segmentation: TMCs are typically segmented at each intersection, while HPMS data is either one complete route or segmented by changes in Data Items. This necessitates many-to-many matches between TMC and HPMS segments.
• Different Representation of the Same Feature: Divided highways, ramps, and intersections may be represented differently not only between INRIX and HPMS but also among different State HPMS data submissions. This requires slight adjustments to the logic based on each of the 50+ State HPMS conflation processes.
• No Common Attributes: HPMS data does not include road names, making name matching unavailable to help increase match rates.

1Spatial configured, deployed, and ran its LRS Conflation Solution built on top of 1Spatial COT’s 1Integrate and 1Data Gateway technology to support conflating every State DOT’s NHS network onto the INRIX TMC segments. The solution takes both datasets and generates an output file identifying the AADT information onto the unique ID of the TMC segment.

The automation processing steps:

• Validation: This initial validation step examines the submitted data in isolation. It begins with a set of validation rules to ensure the data conforms and is fit to proceed through the conflation process.
• Standardization: This step takes inputs from multiple contributors with potentially different formats, schemas, and projections, transforming them into a single format, schema, and projection for processing. By standardizing first, the need to configure matching and conflation logic for every implementation is removed, creating a more repeatable process.
• Change Detection/Matching: This step compares two datasets to identify where they match or differ. Since datasets are often created by different organizations and at different scales, fuzzy spatial and attribution tolerances are required to ensure proper matching.
• Application: This step takes the matches and generates an output file identifying the AADT information onto the unique ID of the TMC segment.
• Final Validation: This final validation step examines the results from the Application process, applying validations to ensure the results meet the needs and are fit for delivery. It identifies gaps in mismatched features in the TMC events, pinpointing spots where manual intervention is required. The system also provides matching metrics to understand how well the conflation process performed.

The automated process successfully conflated 384,803 unique TMCs, representing approximately 131,000 road centerline miles in about 20 weeks. Only 6% of the TMCs required review, effectively automating 94% of the conflation effort. The biggest bottleneck in the process was the manual review step, which was mostly handled by a single FTE person. With additional FTE resources, the conflation could have been completed in a few months. Additionally, this automation significantly reduced subjectivity, resulting in consistent results year over year. Consistency is crucial, as small inconsistencies in conflation can disrupt travel time and performance metrics. Key benefits include:

• Allows CATT Lab to integrate the HPMS NHS data onto INRIX TMC’s to create the NPMRDS product
• Very limited manual intervention required
• Completed conflation task much quicker than previous conflation efforts
• Provides consistent results year over year
• Can conflate and manually address outliers of the entire network in a few months