Truck identification on freeways using Bluetooth data analysis
conference paper
Bluetooth technology is receiving more and more attention to support travel time measurement for intelligent transportation
systems (ITS) applications. Bluetooth receivers are used to time-stamp passing identical vehicles at different locations based on
their unique MAC addresses. This information is useful to predict travel times and estimate origin-destination flows on freeways.
However, there is more valuable information in this big data source than has been explored to date. The main objective of this
paper is to show vehicle type as a new feature that can be extracted from Bluetooth data, presenting a semi-supervised learning
methodology which can be used to identify trucks in freeways. In this paper we also address how to deal with outliers in the
Bluetooth data using an unsupervised machine learning technique to make vehicle identification and other data analysis more
reliable. The predominant application for this vehicle identification is to predict travel time and estimate origin-destination
specifically for freight transport. We use the A15 freeway in the Netherlands as a testbed. This corridor connects the port of
Rotterdam to its hinterland and is one of the important freeways for logistic trip planning. The results show that the proposed
method can identify trucks next to passenger cars with acceptable certainty and improved accuracy.
systems (ITS) applications. Bluetooth receivers are used to time-stamp passing identical vehicles at different locations based on
their unique MAC addresses. This information is useful to predict travel times and estimate origin-destination flows on freeways.
However, there is more valuable information in this big data source than has been explored to date. The main objective of this
paper is to show vehicle type as a new feature that can be extracted from Bluetooth data, presenting a semi-supervised learning
methodology which can be used to identify trucks in freeways. In this paper we also address how to deal with outliers in the
Bluetooth data using an unsupervised machine learning technique to make vehicle identification and other data analysis more
reliable. The predominant application for this vehicle identification is to predict travel time and estimate origin-destination
specifically for freight transport. We use the A15 freeway in the Netherlands as a testbed. This corridor connects the port of
Rotterdam to its hinterland and is one of the important freeways for logistic trip planning. The results show that the proposed
method can identify trucks next to passenger cars with acceptable certainty and improved accuracy.
TNO Identifier
981739
Source title
World Conference on Transport Research - WCTR 2019 Mumbai 26-31 May 2019
Pages
1-11