WS#12 AI in Robot Safety: Methods, Tools, Requirements, and Standardization
Magnus Albert (SICK AG), Saeed Abdolshah (KUKA GmbH), José Saenz (Fraunhofer IFF), Björn Matthias (ABB), Debora Clever (ABB & TU Darmstadt), Michael Rathmair (Joanneum Research), Paolo Barattini (Standardisation TG), Roland Behrens (Fraunhofer IFF), Patrick Courtney (Laboratory Robotics TG), Volker Schmirgel (KUKA GmbH)Room 22-24
Questions to be answered
This workshop offers a unique opportunity to explore the cutting-edge intersection of AI, safety and robotics, a critical and rapidly evolving area. We will dive deep into the challenges and solutions around integrating AI methods into safety functions, providing not only a review of foundational concepts but also discussion on applying AI to safety systems. This workshop is ideal for professionals looking to stay ahead in implementing AI-driven safety in robotics, providing insights into both the current landscape and future developments in safety standardization.
In addition, this workshop brings a unique opportunity to learn from the experience of experts in the automotive industry, a pioneer in applying AI to safety-critical systems. Attendees will hear firsthand about best practices and real-world challenges in the use of AI for safety, with insights directly applicable to robotics. Beyond expert presentations, participants will also engage in a dynamic, interactive panel discussion and network with fellow professionals, fostering collaboration and shared learning in this fast-growing field.
Description
The workshop will start with a brief introduction followed by presentations from three experts who will discuss different aspects of AI and safety, sharing their insights and experiences. These talks will cover key topics to enhance understanding of AI in robotic safety, and essential terminology such as safety, adaptive behavior, pre-emptive behavior, and artificial intelligence. The presentations will also include an overview of safety fundamentals, focusing on the key pillars of safety functions—hardware, software, and potentially data as an additional element for AI-driven methods. Topics such as simulation, AI safety, and AI applications in the automotive industry—and the lessons robotics can draw from them—will also be covered.
Following the talks, a panel discussion will explore critical questions on how AI techniques can integrate with existing safety frameworks and what new features are necessary for AI's successful incorporation into safety practices. The discussion will also touch on the Safety of the Intended Functionality (SOTIF) and functional safety, highlighting the distinction between functional deficiencies and system malfunctions in robotics. Though initially developed in the automotive industry, these concepts could be highly valuable for robotic systems with AI integration. Additional points of focus include the requirements for high-quality datasets, methods to assess their reliability, and the lifecycle of data-driven models used in safety functions. Verification and Validation of these models will be a key topic, along with discussions on how data-driven approaches could shape the future of robotic safety.
Organisation of the WS
We will bring together experts from the fields of robotics, AI, and safety to foster discussions on specific topics, addressing the challenges within each area and using the experience from various disciplines for AI in robot safety. The session will feature three brief talks (each for 15 minutes) from experts in safety, AI, and the automotive industry. The automotive sector, being a pioneer in the adoption of AI for its applications, offers insights that could be particularly valuable and interesting for the robotics community. In addition, a panel discussion/fishbowl discussion will be organized to actively engage attendees, providing them with the opportunity to contribute to the conversation, share insights, and interact with experts (45 minutes). This format will encourage meaningful dialogue, allowing participants to discuss the topics in-depth and exchange perspectives across various disciplines.
.
Intended outcome
- Gain experience from different fields to be applied in robotics
- Form a network and strengthen TG safety, TG industrial robotics, TG Standardisation and TG Laboratory Robotics connections for future discussions and events
- Plan writing a white paper starting with harmonizing AI vocabulary
Speakers, Panelists, etc
Moderators: Saeed Abdolshah, José Saenz
Speakers:
- Prof. Simon Burton, University of York
Title: AI Safety: What can robotics learn from automotive industry
- Dr. Magnus Albert, SICK AG
Title: Future of Robotics: Simulation & aspects of AI Safety
- Dr. Björn Matthias, ABB
Title: Data-driven methods challenges for safety
Topic Groups and/or Innovation networks involved
TGs: Safety, Standardisation, Industrial Robotics, Laboratory Robotics
Projects involved
NA
Further information
One paper from each of speakers
Organisers
- Magnus Albert (SICK AG), magnus.albert@sick.de
- Saeed Abdolshah (KUKA GmbH), saeed.abdolshah@kuka.com
- José Saenz (Fraunhofer IFF), jose.saenz@iff.fraunhofer.de
- Björn Matthias (ABB), bjoern.matthias@de.abb.com
- Debora Clever (ABB & TU Darmstadt), debora.clever@de.abb.com
- Michael Rathmair (Joanneum Research), michael.rathmair@joanneum.at
- Paolo Barattini (Standardisation TG), paolo.barattini@yahoo.it
- Roland Behrens (Fraunhofer IFF), roland.behrens@iff.fraunhofer.de
- Patrick Courtney (Laboratory Robotics TG), patrick.courtney@tec-connection.com
- Volker Schmirgel (KUKA GmbH) volker.schmirgel@kuka.com