Fraunhofer Institute for Cognitive Systems IKS
Fraunhofer Institute for Cognitive Systems IKS

Trustworthy autonomous systems

Two Fraunhofer IKS researchers are participating in the »European Training Network for Safer Autonomous Systems« project, which is being funded within the framework of the EU »Horizon 2020« innovation initiative. The goal of the pan-European research project is to come up with safety strategies for all stages of autonomous system development. The IKS researchers are focusing their efforts on designing inherently safe software architectures.

Privacy warning

With the click on the play button an external video from www.youtube.com is loaded and started. Your data is possible transferred and stored to third party. Do not start the video if you disagree. Find more about the youtube privacy statement under the following link: https://policies.google.com/privacy

MSCA ETN SAS - European Training Network for Safer Autonomous Systems

Safety that is already anchored into the software design

This project involves supporting and training elite young scientists in the area of safe autonomous systems. As part of their joint research, the select researchers will be tasked with developing a trained eye on autonomous systems with a special focus on safety. Using a holistic approach, the project is structured around successive sub-projects that stretch across the entire innovation path, from the development of robust software architectures and continuous safety management, to application in concrete case studies. The central goal of the project is to develop safety strategies for autonomous systems so that users can place their trust in the technologies.

Inherently safe systems

Both of the Fraunhofer IKS researchers are working on the development of inherently safe autonomous systems. The only way to guarantee the safety of intelligent systems is through this safety-by-design approach. By integrating high-performance validation mechanisms directly into the design of the cognitive systems, manufacturers and users can put their trust in the safe functionality of the system.

Safety by Design

The Fraunhofer IKS researchers are handling the real-time adaptation and error tolerance, and building on top of that dynamic safety handling at runtime via real-time safety contracts.

Real-time adaptation and error tolerance mechanisms

While real-time adaptation enhances the autonomous capabilities of intelligent systems, the overall system is subject to new safety risks through the adaptive software architectures. And although conventional validation methods assume worst-case conditions, they place extremely stark limitations on the degree of freedom of the system as a result. With this in mind, Fraunhofer IKS researchers are developing dynamic adaptation strategies. Given that unsafe middleware can be a source of errors,  young researcher Yuan Liao is developing a platform for safe adaptive systems. This platform detects safety-critical software components and automatically switches to a safe configuration as a means of automatically adapting to the higher safety requirements. When designing new software platforms, many developers first improve performance through markers such as the data rate, latency and memory consumption. Yuan Liao and his colleagues at Fraunhofer IKS are working to simultaneously improve the robustness and reliability of the systems. This degree of safety, even under critical internal or external conditions, must be verifiable, which the researchers are achieving through model-based analysis techniques.

Dynamic safety handling through runtime safety contracts

The goal of the second sub-project is to design validation strategies that combine design mechanisms (such as adaptive software architectures) and information from data that the system collects. After all, autonomous systems learn and continuously develop. For this reason the validation must be viewed as an on-going process. When it comes to validating the system, for one thing Fraunhofer IKS researcher Joao-Vitor Zacchi  does not want to lose sight of the performance. Secondly, the safety of the interacting autonomous systems cannot be viewed on an isolated basis. The safety of the overall traffic situation in an autonomous vehicle system can be improved only if the individual vehicles communicate with one another and receive further information on their surroundings as a result. Fraunhofer IKS researchers are implementing this concept via real-time safety contracts. The individual systems jointly agree on which boundaries the individual systems are not permitted to overstep and then coordinate their behavior. This is similar to simple hand signs that human traffic participants use to understand one another and coordinate their behavior.

Marie Skłodowska-Curie Actions

Yuan Liao’s and João-Vitor Zacchi’s research activities will be funded through a grant from the Marie Skłodowska-Curie Actions Research Fellowship Program, which promotes young, highly-engaged researchers. As select members of the European Training Network for Safer Autonomous Systems, research fellows receive support during the early phase of their careers and are networked with experts and other talented young researchers. Business and research institutions hire these young scientists and support them through the exchange of ideas and the opportunity to apply their solutions in practical applications.

This project has received funding from the European Union's EU Framework Programme for Research and Innovation Horizon 2020 under Grant Agreement No # 812788