FUSE: FUnctional Safety and Evolvable architectures for autonomy
How can self-awareness be introduced into a machine, so that the machine can operate itself?
This is the question that motivates us. We are examining the ways and means by which machines can be designed, so that their operation no longer requires active human supervision. In effect, these machines would be AUTONOMOUS. Within the FUSE project, we focus on a particular category of machines: The Self-driving Vehicle, and in particular investigate appropriate architectures and how to ensure safety of such self-driving vehicles.
Why self-driving vehicles and FUSE?
Self-driving vehicles add a number of desirable properties to transportation:
- Self-driving vehicles prevent and perform better in accident scenarios. Therefore, they make transportation safer.
- Automatic coordination among self-driving vehicles can increase the number of vehicles that efficiently operate within an area. Therefore, the traffic capacity increases.
- A self-driving vehicle or fleet of vehicles can constantly operate in the most fuel efficient manner, thus reducing environmental impact as well as operational costs.
- [Also, self-driving vehicles are COOL and designing them is A LOT of fun! But of course, that is not why we are doing it ;-) ]
Before self-driving cars can be allowed on public roads, a lot of work needs to be done to ensure that they are safe. The FUSE project therefore has two complementary objectives:
- Specify how the 'self-driving' part should be designed in self-driving vehicles and
- Develop ways to ensure and demonstrate the safety aspects of self-driving vehicles
Read on for a more formal project description.
Duration: Q4 2013 to Q3 2016.
The FUSE project focuses on system architectures and functional safety, from the perspective of system autonomy. Autonomy has many facets and represents a paradigm shift for the automotive industry. Current automotive systems and functional safety standards are evolving, but so far they have only considered autonomy to a limited extent. Therefore, investigations are urgently needed into the limitations of the existing sytsems and the ISO26262 safety standard. Propositions for overcoming the limitations are also needed. The functional safety considerations and the scalability and cost-efficiency of architectures represent potential blocking factors for introducing new autonomy functions. The FUSE project aims to
- Develop an understanding of limitations of current systems and safety standards.
- Develop requirements for autonomy considering safety and architecture.
- Create guidelines and update suggestions for ISO26262.
- Create a reference architecture.
- Validate the solutions in realistic vehicle set-up
The work within FUSE is split into multiple work packages (WPs) as shown in the figure above. Each WP is lead by one of the consortium partners and supported by other partners to a certain extent.
KTH is the leader for WP3 (Architecture) as well as an active participant in WP1 (Evaluation in Realistic Vehicle Setup) and WP4 (Networking and Dissemination).
We at KTH have a strong research background in architecting new and innovative automotive systems. For example, the provided an early theory and implementation of dynamically self-reconfiguring automotive systems. The Scoop project developed a reference architecture and implementation for autonomous, cooperative driving. In addition to the architecture itself, we also conduct significant research in representing and reasoning about architectures. In fact, KTH is one of the developers of the architecture description language EAST-ADL . Leading the Architecture work package within FUSE is, therefore, a promising extension of our ongoing research.
The KTH team within the FUSE project has strong engineering talent and a keenness for applied research. Therefore, we have taken an active interest in WP1 which involves the application of our research in a realistic vehicle setup. This way, our research results have the potential to make it into real products, in addition to academic publications.
Finally, as a university, KTH exists to increase and spread knowledge. Therefore, we will naturally be involved in WP4 - the networking and dissemination of our results. This would be in the form of publications as well as hosting and participating in workshops, seminars and conferences.
KTH:s planned contribution
The main contributions have concerned architectures for autonomous vehicles. Current work addresses safety for perception related functionality.
When developing a 'new and improved' architecture, it is necessary to first understand the requirements of the new architecture as well as the short-comings of the existing architecture. Therefore, our first contribution would be
- A gap analysis, identifying bottlenecks in existing architectures and the desirable properties of the autonomous architecture to be designed
Once the gap analysis is done, it is time to propose a new architecture, which consolidates patterns, principals as well as guidelines for application. Thus, the second contribution would be
- A reference architecture for autonomous vehicle propulsion
Finally, there would be a need to lay a clear path to close the gap between existing and newly proposed architectures. Therefore, the third contribution would be
- Strategies for evolution and migration of existing architectures towards autonomy
These contributions will be realized via a number of publications, workshops and other artifacts.
- Architecture challenges for intelligent autonomous machines
Behere S., Asplund F., Söderberg A., Törngren M.
13th international conference on Intelligent Autonomous Machines (IAS-13), Padova, Italy, July 2014
- For further publications, see the recent PhD thesis by Sagar Behere
Architectures for Autonomous Machines
Workshop at KTH arranged by ICES and the FUSE project on 14 January 2014
For further workshops organized in 2015, see events and previous events on the ICES home page: www.ices.kth.se
- Slava Izosimov