Language-Grounded Robot Group Joining Shows Real-Time Grounding and Multimodal Poses
A robot uses a natural-language instruction to identify a target group and predicts where to join in a socially acceptable way.
- Publication
- arXiv
- Stage
- Preprint
- What we read
- Summary of the paper
- Authors
- Zilin Fang, Zishuo Wang, Gim Hee Lee, David Hsu
- Universities and research institutions
- Not yet supplied in verified metadata; the Brief does not guess.
What they did and found
Researchers defined a two-stage approach: ground language to a target group in view, then predict a distribution of feasible joining poses. In tests with conversations, queues, and audiences, the method achieved near real-time grounding and multi-modal joining poses, outperforming baselines.
Why it matters
Understanding how language-guided grouping works helps robots navigate people more safely and naturally, without assuming a fixed goal.
The grounding stage builds candidate groups from the scene and scores them against the instruction. It then maps the chosen group into a top-down representation that guides where the robot stands and which direction it faces.
Real-robot tests show the robot joining standing conversations and queues with collision-free paths, updating its goal when the group changes, though grounding errors can still limit some results.
Original sources · 1
Check the original paper for its authors, methods, version and access terms.
What remains uncertain
Grounding errors and multimodal joining options can still lead to distant or invalid poses in complex scenes.
What is your take?
Ask a question, add useful context or share a different perspective. Keep the conversation respectful and grounded.