Quality analysis of a transport mode detection service using fuzzy rules

Kurzfassung

1 Objectives (150) Since decades for many applications in traffic management and mobility services mobility patterns, movement profiles and traces of people are of big interest. This reaches from origin destination matrices, with which the demand for long- and mid-term traffic simulation models is defined and typically fed by results from surveys, to traffic information services using traces of vehicles as basis for current traffic state detection (Floating Car Data). The vast dissemination of smartphones over the last years with sensors like GPS and accelerator data is a great possibility to substantially enhance the quality of according models and applications. In most cases the determination of the used transport modes (CAR, TRAIN, BIKE, …) is a crucial thing to do, as far as possible in an automated way. Some algorithms already exist, typically focusing on the detection of the used transport modes on single trips, while algorithms for multi- or inter-modal are rare. 2 Data and methodology (150) This contribution presents an analysis on the quality of such a multi-modal method. The basics of the algorithm using fuzzy rules and limiting the algorithm to the use of only the sensor data and nothing additionally (e.g. infrastructure, general traffic situation) are already published, but will be briefly sketched in the introduction of the paper. An enhancement of it will be presented especially concerning the determination of road traffic to raise the quality of jam detection which supplies mobility information systems. The quality analysis is done by analyzing some hundred trips for which the transport mode CAR is known serving as a ground truth. 3 Expected results (100) The quality of the output are shown exemplarily for some trips using map visualizations, diagrams of time series for selected variables (speed, acceleration-sensor, …) and for the whole data set frequency and error measures and distributions. The authors expect quality being high enough to set up a service to which anyone may send sensor data in order to obtain transport modes which may be used by applications not only from the field of transportation, but also other fields of location based services. Such service will also be sketched if it is in an advanced state till the conference deadlines.