How should robots behave when navigating around people? Our framework proposes that robots should adhere to social norms, behave safely, and navigate legibly. We confirm this through an online user study.
Multimodal Augmented Reality for Composing and Executing Robot Tasks (MARCER) is a multimodal end-user robot programming system that combines large language models, trigger-action programming, and augmented reality to enable users to create reactive robot programs using verbal commands. This was published at HRI 2025.
Our system leverages augmented reality to simplify demonstration capture and automatically expand a single human demonstration into a set of synthetic demonstrations, minimizing user effort and collection time.
A Best Paper Honorable Mention Award! We designed and evaluated an immersive Cyber-Physical Control Room interface for remote mobile robots, showing it improved operator effectiveness by 28% in a mock disaster response scenario and enhanced human-robot teamwork.
