Assessing the performance of the SpeedInfo sensor.

PDFTech ReportFHWA/OH-2013/21State Job Number 134723FreewaysTraffic dataData collectionPerformance measurementLoop detectorsSensorsTraffic surveillanceOhioOhio. Dept. of Transportation. Office of Statewide Planning and ResearchCoifman, BenjaminKim, SeoungbumOhio State UniversityUnited States. Federal Highway AdministrationUS Transportation CollectionUntil recently freeway traffic operations data were collected in house by the Ohio Department ofTransportation (ODOT) using loop detectors. In recent years several private companies have emerged with aviable opportunity to outsource traffic data collection. While these new business models and ease ofdeployment are compelling, there are few published studies explicitly evaluating their performance. ODOT hascontracted with two of these data providers: SpeedInfo and INRIX. This study exploited a unique juncture asODOT transitioned to the new third party data collection: the two new systems and the legacy loop detectorswere concurrently operational in Columbus, OH for approximately 6 months. Using the archived data from thisperiod, we evaluated SpeedInfo and INRIX performance over a 14 mi. This long period of time ensured that wewould have a better chance of observing intermittent features that might go unobserved in a short-term study.Both SpeedInfo and INRIX traffic data generally performed within specifications, exhibiting small biasesthat could likely be fixed with fine-tuning. Both sensors also exhibited a reporting lag, on the order of 60 sec forSpeedInfo and 360 seconds for INRIX (comparable to conventional 60 sec and 5 min aggregation periods usedin loop detectors). Although the lag and bias are within specifications, some applications might be sensitive tothem, e.g., traffic responsive ramp metering. Each system also exhibited unique behavior not found in the other.SpeedInfo is sensitive to precipitation: sometimes going offline for tens of minutes. When mounted on theshoulder, SpeedInfo showed slightly degraded performance on the far side traffic. INRIX reported speeds everyminute, but after excluding repeated values, the actual effective reporting period was more like 3-5 min, withoccasional periods of repeated measurements lasting in excess of 10 min (already reflected in the lag above).INRIX also provides two measures of confidence, neither of which reflects the measurement repetition