program

Full-day, in-person program on May 24 or 28, 2027. Room to be announced.

Each invited talk is 25 minutes followed by 5 minutes of questions. Coffee breaks double as poster time.

Time Session
08:30–08:45 Welcome and introduction
The two tracks and the shared research questions.
08:45–10:15 Track 1 invited talks: adapting pretrained general-purpose policies
Shuran Song (Stanford University)
Andreea Bobu (MIT)
Abhishek Gupta (University of Washington)
10:15–10:45 Coffee break and posters
10:45–11:15 Bridging talk
Georgia Chalvatzaki (TU Darmstadt): Mobile manipulation
11:15–12:00 Contributed spotlight talks
Up to nine spotlights, five minutes each.
12:00–12:30 Poster session
Up to 30 posters.
12:30–14:00 Lunch
14:00–15:30 Track 2 invited talks: continual learning from experience
Lukas Rosenberger Schmid (TU Nuremberg)
Hae-Won Park (KAIST)
Sorina Lupu (NASA JPL)
15:30–16:00 Coffee break and posters
16:00–17:45 Deployment cases and panel discussion
Four panelists present real-world deployment cases, followed by two organizer-moderated discussion rounds with the audience.
Pascal Klink (Amazon Robotics)
Aykut Onol (Toyota Research Institute)
Abrar Anwar (NVIDIA)
Nico Bohlinger (TU Darmstadt & Amazon FAR)
17:45–18:00 Synthesis, Best Poster Award, and closing

More about the invited speakers is on the speakers page.

Deployment cases and panel discussion

The afternoon session combines the panel and the community discussion into a single case-based session. Each panelist briefly presents a real-world case of continual robot learning, including its practical constraints, failures, and lessons learned. Two organizer-moderated rounds then open the discussion to the audience, organized around the workshop’s shared research questions.

The session ends with a group synthesis of the main challenges, open questions, and evaluation priorities. This synthesis will inform a research agenda that the organizers will publish after the workshop.

After the workshop

Slides and any available recordings will be shared here afterwards, with the consent of the speakers.