The FAILSAFE project has significantly enhanced our understanding of human-robot interaction (HRI), particularly in how users define failures and assess risks tied to anticipated failures. In collaboration with our industry partner Beko Plc. we explored the concept of failures in a domestic HRI setting, focusing on a dishwasher robot. Over the course of our year-long research, we revisited existing failure taxonomies and applied them within the context of domestic environments, uncovering social and corporate dimensions of failure that are often neglected. A key output is our new taxonomy of failure outcomes, which highlight the perceived risks that such breaches pose to the overall HRI experience and how failures can erode trust. Insights from four FAILSAFE workshops have been disseminated through a poster presentation, a full paper and media coverage.
As the FAILSAFE project concludes, we’re proud of what we’ve achieved and excited for the next phase of research. Our continued work will push the boundaries of human-robot interaction, ensuring that robots can handle failures more effectively and improve their resilience in real-world scenarios. Through our new collaborations, we’re committed to advancing the field of HRI and making robots smarter, safer, and more capable of learning from their mistakes.
Key Findings
Throughout FAILSAFE, we conducted four workshops with both expert and non-expert stakeholders. Workshop participants highlighted that failure in domestic robots impacted trust differently depending on the type of failure and its source. Catastrophic failures, which are often attributed to social- or service provider-related issues, severely damaged relation-based trust, while moderate failures primarily affected performance-based trust, making users more cautious. Interestingly, our studies revealed that the most significant failures were those the users couldn’t easily understand, emphasizing the importance of transparency and clear communication in robot design.
Research Contributions
Poster and Paper Publications:
Our findings were shared through a poster presented at the International Conference on Robotics and Automation (ICRA’23) and a full paper at the International Symposium on Trustworthy Autonomous Systems (TAS’24).
Media Coverage:
FAILSAFE was also featured in Robot Talk Episode 63, providing an excellent opportunity to share our work with a broader audience. If you haven’t had a chance to listen, you can find it here (23:00 in).
Looking Ahead
Our team is actively preparing a journal submission, expanding our focus on how failures propagate within broader human-robot ecosystems. This new direction will allow us to investigate failure and recovery across complex interaction networks, ultimately redefining robot resilience in dynamic environments.
The Next Chapter
Our success in FAILSAFE has led to additional funding through the EPSRC UK-RAS Network. The new project, Principles of Learning from Unstructured Human-Robot Interactions, in collaboration with the University of Lincoln, will focus on real-world HRI data generated by Lindsey, a robot tour guide operating at the Lincoln Collection Museum since 2018 as part of a long-term partnership with Lincolnshire County Council. This research will explore failure modes and how robots can learn from unstructured interactions, where human behaviour is complex and unstructured. A key takeaway is the importance of not just identifying failures but learning from them to improve future interactions.
To further this research, we are also excited to welcome Raul Ghisa as a new EPSRC CASE PhD student in the School of Computer Science at the University of Nottingham. His PhD project on Failure Understanding and Recovery in Robotics will build on the work we’ve begun, helping us develop methods for robots to handle failure scenarios gracefully, detect and recover in real-time, and learn from both failures and human assistance. This work will be crucial in enhancing resilience in human-robot ecosystems and ensuring that robots continue to improve their interactions in complex environments.
Horizon Digital Economy Research, University of
Nottingham Innovation Park, Triumph Road, Nottingham,
NG7 2TU
horizon@nottingham.ac.uk
