In English

Detecting roundabout manoeuvres using OpenStreetMap and vehicle state

Fernando Jorge
Göteborg : Chalmers tekniska högskola, 2012. Diploma work - Department of Applied Mechanics, Chalmers University of Technology, Göteborg, Sweden, ISSN 1652-8557; 2012:54, 2012.
[Examensarbete på avancerad nivå]

Naturalistic Driving Studies (NDS) is a recent research eld used for the analysis of behaviour of road users in a natural setting. NDS provide large data sets containing numerous types of events such as manoeuvres, near crashes and crashes. Since we are dealing with a large database the analysis of those events can be hard due to the lack of information of where or when those events occur in the trips data. The identi cation of these events is typically done by a person manually reviewing the video footages. This is a tedious, time-consuming and ine ective way of identifying events in a large database. This project aims for the development of a method for an automatic identi cation of some car manoeuvres(i.e., left manoeuvre in a roundabout, right manoeuvre in a roundabout, and going straight in a roundabout) as well as a speci c road structure(i.e., roundabouts). The data for this study comes from a Field Operation Test done in Gothenburg, as part of the euroFOT project. Some information from the OpenStreetMap project was also used. The labelling method developed here consists of the creation of a yaw rate pro le for each manoeuvre under study that is then matched to the trip segment being analysed through the use of the Pearson's correlation parameter. The roundabout identi cation utilizes the results from the labelling method in order to gather the localization (GPS coordinates) of those roundabouts. The results obtained indicate that the method developed here has potential to be used for an automatic way of identifying speci c roundabout manoeuvres that exist in a large database as well as the identi cation of the localization of roundabouts.

Nyckelord: Sensor fusion, Field Operational Test, Driver behaviour, Machine learning



Publikationen registrerades 2013-01-16. Den ändrades senast 2013-04-04

CPL ID: 170939

Detta är en tjänst från Chalmers bibliotek