To the top

Page Manager: Webmaster
Last update: 9/11/2012 3:13 PM

Tell a friend about this page
Print version

How autonomous driving in… - University of Gothenburg, Sweden Till startsida
Sitemap
To content Read more about how we use cookies on gu.se

How autonomous driving influences the vehicle's architecture

Conference paper
Authors Christian Berger
Published in 2016 Workshop on Automotive Systems/Software Architectures (Wasa). 5-8 April 2016
ISBN 978-1-5090-2571-8
Publisher IEEE
Publication year 2016
Published at Department of Computer Science and Engineering (GU)
Language en
Links dx.doi.org/10.1109/WASA.2016.12
Keywords autonomous driving, vehicle architecture, self-driving vehicle, Computer Science, Transportation
Subject categories Transport Systems and Logistics, Computer and Information Science

Abstract

Since the 2007 DARPA Urban Challenge, the world's largest experiment with driverless vehicles having to safely interact with each other in a fenced area imitating an urban environment, all important vehicle manufacturers have established research and development around this technology to make driving safer and more comfortable for their customers. The functionality therefore is realized by software outgrowing all individual software units currently present in today's cars to process data volumes, currently in the range of several hundreds megabytes per second from the sensors perceiving a vehicle's surrounding, to derive safe driving decisions. This talk gives an overview of an interdisciplinary research project at Chalmers University of Technology and University of Gothenburg envisioning "CampusShuttle", a self-driving vehicle tackling inner-city driving scenarios. An outline is given for the challenges arising from the embodied technology on the vehicle's sensor architecture, its computing architecture to process the data provided by the on-board and off-board data sources, and its software architecture. Besides these fundamentally necessary aspects, this technology's impact on the test and delivery architecture for the software units as well as the operating deployment architecture is presented and discussed. As the algorithms enabling this technology are predominantly data-driven by the volatile vehicle's driving scenarios, a complete and consistent requirements specification for the building blocks in the software is hardly possible to start with; complementary thereto, recent trends evaluate algorithmic approaches that use growing amounts of data from different driving scenarios in machine learning. In such a highly evolutionary development context, where the software needs to be adapted to better meet requirements that are continuously unveiled during testing and operation, continuous integration of the individual software units comprising a self-driving vehicle system, fast continuous delivery of the resulting executable artifacts to the self-driving vehicle, and their safe and reliable continuous deployment thereon is crucial to prepare experiments, evaluate and learn from the collected data, and to traceably maintain the software environment.

Page Manager: Webmaster|Last update: 9/11/2012
Share:

The University of Gothenburg uses cookies to provide you with the best possible user experience. By continuing on this website, you approve of our use of cookies.  What are cookies?