Validation of TomTom historical average speeds on freeway segments in Gauteng, South Africa

[摘要] ENGLISH ABSTRACT: Traditional methods of traffic data collection, such as inductive loops and road sensors, continue to be the main source of traffic data. The advancement in technology and vehicle tracking methods has proved to be the impetus behind the emerging of alternative and innovative sources of traffic data, such as ITS data sources. ITS sources, such as vehicle probes, are becoming increasingly important due to their low cost and the vast amounts of traffic data produced. However, traffic data from ITS sources raise new concerns about data quality. The quality of probe data in South Africa and other developing countries is unknown. This study sets out to investigate the quality of TomTom historical average speeds on selected freeway segments in South Africa.The study compared TomTom historical speed estimates and reference speeds on six directional segments on the N1 and R21 freeways. The reference data used was Automatic Number Plate Recognition (ANPR) data, a component of Open Road Tolling (ORT) in Gauteng. A freeway segment is the road section between two toll gantries. All 15-minute and 1-hour intervals between 05:00 and 20:00 during the weekdays (Monday – Friday) in February 2015 were grouped and aggregated. The quality measures evaluated were accuracy, completeness, validity, coverage and accessibility.To evaluate accuracy, three error quantities were determined, namely signed error, average absolute speed error (AASE) and speed error bias (SEB). The allowable errors for the signed error, AASE and SEB were ±10 %, 10 km/h and ±7.5 km/h, respectively. TomTom speeds were highly consistent with the reference speeds. The error quantities for the combined freeway segments were less than the allowable errors. The signed errors and AASE for all the six individual freeway segments were also less than the allowable errors. In five of the six sections, the SEB was less than the allowable error. There were no significant differences between the error quantities derived from 15-minute and 1-hour interval speeds for the combined and individual freeway segments. On the other hand, validity was dependent on the selected measure. TomTom speeds were of very high quality based on the signed error and AASE, whereas the same data was of moderate quality based on the SEB.Although the TomTom speeds were within the specified accuracy thresholds, the speed estimates were generally lower than the reference speeds throughout the analysis period. TomTom estimates were better at low speeds and the quality of TomTom estimates declined with an increase in speed. It is possible that the low TomTom speed estimates were due to a sample that was not a true representation of the traffic stream. Importantly, it is possible to enhance the accuracy of TomTom speed estimates by using certain percentile speeds instead of average speeds.

[发布机构] Stellenbosch University