| dc.description.abstract | Historical Traffic Pattern (HTP) information may be used in automobile Personal Navigation Devices (PNDs) to more accurately predict the estimated time of arrival (ETA) at a destination, along a route defined by the PND. The objective is to compare two HTP data sources that Enterprise X could utilize. “Source A” is available for royalty fees. In addition, Probe Vehicle Data (GPS location at time intervals that are variable, typically between 1 and 60 seconds) is available from “Source B”. It is conceivable to compile this Source B probe data and statistically analyze it in order to formulate an HTP model. If research can show Source B HTP data quality to be comparable or superior to Source A in a given sample region, it would support an Enterprise X decision to pursue using Source B HTP data in Europe. In the project, HTPs are considered as a collection of point estimates of speed. The speed sample statistics (mean, standard deviation, and sample size) are used to calculate a probability or quality-of-fit metric for a given speed estimate. An accepted method of combining probabilities is then applied to calculate a composite quality-of-fit statistic for a candidate Standard Traffic Pattern. In order to exercise the model and compare its efficacy to that of existing Source A, the procedure is to traverse actual ground truth trip data aside from that which was used in the development of the Source B model. Use Travel Time Error (TTE) to score the ground truth data against the predicted ETA using (a) the Source A HTP data, and (b) the Source B HTP data. Results indicate that the experimental Source B RMS TTE statistics versus trip distance follow a similar pattern to the established Source A data source, with Source B having slightly larger RMS error for shorter trip distances, particularly shorter than 8 miles. Regarding optimistic vs. pessimistic prediction bias (optimistic bias is where the estimated travel time is shorter than the actual travel time from ground truth data), Source A has a generally pessimistic bias, with the exception of shorter trips, less than 4 miles. Source B has a more optimistic bias than source A, across the span of trip distances. The conclusion is that accuracy of Source B is comparable but not quite as good as that of Source A across the range of trip distances, meeting the requirement put forth in the stated objective. | |
