Error assessment for emerging traffic data collection devices.

PDFTech ReportWA-RD 810.1T4118, Task 46DTRT07-G-00Travel timeTraffic estimationTraffic volumeTraffic speedSensorsData collectionDecision support systemsTraffic dataAlternatives analysisAccuracyErrorsReliabilityCostsWashingtonWashington (State). Dept. of TransportationWang, YinhaiAraghi, Bahar NamakiMalinovskiy, YegorCorey, JonathanCheng, TianxingPacific Northwest Transportation ConsortiumUnited States. Dept. of Transportation. Office of the SecretaryUS Transportation CollectionBecause access to travel time information can significantly influence the decision making of both agenciesand travelers, accurate and reliable travel time information is increasingly needed. One important step inproviding that information is to identify the sensors best suited to provide travel time data for a givencorridor. Currently, few studies have evaluated the effectiveness of various travel time data collectiontechnologies side-by-side. This evaluation was intended to provide decision support for transportationagencies looking to select travel time systems on the basis of accuracy, reliability, and cost. This studyfocused on two test corridors: State Route 522 (SR 522) (an urban arterial with frequent intersections)between the NE 153rd Street and 83rd Place NE intersections, and I-90 (rural freeway built overSnoqualmie Pass in the Cascade mountains) from milepost 109 (Ellensburg, Wash.) to milepost 32 (NorthBend, Wash). The sensor systems tested were Washington State Department of Transportation\ue2\u20ac\u2122s pre-existing automatic license plate reader (ALPR) system, Sensys emplacements, the TrafficCast BlueTOADsystem, Blip Systems BlipTrack sensors, and a third-party feed from Inrix. This study\ue2\u20ac\u2122s approach was tolook at the Mean Absolute Deviation (MAD) to judge the expected magnitude of error, then examine theMean Percent Error (MPE) to find any systematic biases in the data. The Mean Absolute Percent Error(MAPE) was useful for finding the relative magnitude of the error, and the Root Mean Square Error(RMSE) was used to determine whether a few large errors or many smaller errors were occurring. Eachsystem in the analysis demonstrated different strengths and weaknesses that should be considered inaddition to its accuracy and sample rates. Some systems can provide additional data; others trade accuracyand coverage for cost or portability. Ultimately, engineers will need to weigh their requirements foraccuracy and sample rates against the other engineering constraints imposed on their system