Workshop Highlights | 2nd Workshop on Robot Execution Failures and Failure Management Strategies at RSS 2024

Our workshop took place on July 15th and we would like to thank all speakers, panelists, presenters of contributed papers, and participants in general for making it a great success! It was particularly nice to have a variety of different lines of work and perspectives represented at the workshop, illustrating the very valuable developments that are being done to mitigate the effects of robot failures in human-robot interaction and collaboration. The main highlights of the workshop are presented below in a visual summary.

Invited Talks

We had six very interesting invited talks at the workshop (each followed by a Q&A session), by: Prof. Shuran Song, Prof. Dongheui Lee, Prof. Chris Pek, Alessandra Sciutti, Ph.D., Prof. Maartje de Graaf, and Mr. Maximilian Diehl. More details about the speakers can be found on the speakers page.

Panel Discussion

In the panel discussion, we had six panelists (Prof. Shuran Song, Prof. Dongheui Lee, Prof. Carlos Hernandez Corbato, Alessandra Sciutti, Ph.D., Prof. Maartje de Graaf, and Maximilian Diehl), who had an interesting discussion on the underlying topic "Challenges in handling execution failures in human-robot interaction and collaboration." Various points were touched upon during the discussion, such as:

Whether the advancements in LLMs are sufficient for resolving robot execution failures in general?
Whether aiming for general robot policies is meaningful if a robot is only supposed to be applied in a constrained domain?
What is the benefit of robot modularity on execution failures and trust in interaction and collaboration?
What role do (psychological) models of interaction play in the treatment of execution failures?
What are the potential challenges of using foundations models with respect to (perceived) failures if a robot is deployed to cultures that are different from the dominant culture represented by the training data?
What are some tips for young researchers who are interested in working on making robots more failure-aware and failure tolerant?

Interactive Session

All participants who were able to stay until the end of the workshop also had a round-table discussion on common types of failures that they have encountered in robot applications, strategies for mitigating those, effects of failures on robot acceptance, as well as the representation of failures in robot learning datasets.

Paper Spotlight Presentations

Three contributed papers were also presented at the workshop, first by a short spotlight presentation, and also in poster presentations during the coffee breaks. The PDFs of the contributed papers can be found on the accepted papers page.