Abstract
Bluetooth detectors are becoming increasingly popular as a technology for acquiring travel time data. However, these data include outliers that are caused by several factors, and a robust outlier detection algorithm is needed for filtering out the outliers. Arterial roadways present a particularly challenging environment because the traffic control devices introduce a large amount of variability to the measured individual travel times and because of the prevalence of other sources of error (e.g., en route stops, Bluetooth-enabled devices not in vehicles). This paper presents a new adaptive outlier detection algorithm that is proactive rather than reactive. Unlike conventional reactive algorithms that rely solely on recent data, the proposed algorithm uses both historical data and current data to predict the validity window. The performance characteristics of the proposed algorithm are illustrated, and field data from a signalized arterial are used to compare the proposed algorithm and a benchmark algorithm. The results show that the proposed model is superior to the benchmark model and that the model performs well across a wide range of traffic conditions.
Get full access to this article
View all access options for this article.