Repository hosted by TU Delft Library

Home · Contact · About · Disclaimer ·
 

Classification for Safety-Critical Car-Cyclist Scenarios Using Machine Learning

Publication files not online:

Author: Cara, I. · Gelder, E.D.
Type:article
Date:2015
Publisher: Institute of Electrical and Electronics Engineers Inc.
Source:IEEE Conference on Intelligent Transportation Systems, Proceedings, ITSC, 2015-October, 1995-2000
Identifier: 530897
doi: DOI:10.1109/ITSC.2015.323
ISBN: 9781467365956
Article number: 7313415
Keywords: Traffic · Accidents · Artificial intelligence · Safety engineering · Machine learning techniques · Advanced driver assistance systems · Mobility · Fluid & Solid Mechanics · IVS - Integrated Vehicle Safety · TS - Technical Sciences

Abstract

The number of fatal car-cyclist accidents is increasing. Advanced Driver Assistance Systems (ADAS) can improve the safety of cyclists, but they need to be tested with realistic safety-critical car-cyclist scenarios. In order to store only relevant scenarios, an online classification algorithm is needed. We demonstrate that machine learning techniques can be used to detect and classify those scenarios based on their trajectory data. A dataset consisting of 99 realistic car-cyclist scenarios is gathered using an instrumented vehicle. We achieved a classification accuracy of the gathered data of 87.9%. The execution time of only 45.8 us shows that the algorithm is suitable for online purposes. cop. 2015 IEEE.