Abstract

How can a robot provide unobtrusive physical support within a group ofhumans? We present Attentive Support, a novel interaction concept for robots tosupport a group of humans. It combines scene perception, dialogue acquisition,situation understanding, and behavior generation with the common-sensereasoning capabilities of Large Language Models (LLMs). In addition tofollowing user instructions, Attentive Support is capable of deciding when andhow to support the humans, and when to remain silent to not disturb the group.With a diverse set of scenarios, we show and evaluate the robot's attentivebehavior, which supports and helps the humans when required, while notdisturbing if no help is needed.

 

Quick Read (beta)