The acquisition of travel data is currently based on cost- and time-intensive questionnaires and yields mostly an incomplete picture due to limited coverage and inadequate updates. There is an urgent need for technologically supported data acquisition tools. This paper introduces a novel approach to developing a large-scale travel survey by intelligently employing data from smartphones. Based on signals of the embedded accelerometers and GPS reveivers, an ensemble of probabilistic classifiers is trained for automatically reconstructing the individual trips composing a tour, including the mode choice. In the region of Vienna, Austria, 266hours of travel data were collected to train and evaluate the models. Using a set of 72 features, the best classification results are achieved for detecting walks (92%) and bike rides (98%). Railway modes were correctly identified in 80% of all cases, which is subject to further research. In case of GPS losses only accelerometer data are used, which still shows promising results. This allows the method to incorporate places where there is normally only a weak or no GPS signal. Future smartphone applications are intended to spread the tool among traffic users, while the effort for them should be kept to a minimum i.e. no manual entries or questionnaires are necessary. Due to the increasing popularity of smartphones, the tool has the potential to be used on a wide-spread basis.
展开▼
