A Novel Architectural Method for Producing Dynamic Gaze Behavior in Human-Robot Interactions
summary
Integration of DIARC and ARCADIA produces a working pipeline that uses 3D vision from a PR2/Fetch, DIARC one-shot object learning, and ARCADIA's attention cycles to drive gaze via a simple inhibition of return plus a semantic bias from dialogue. The system generates active gaze shifts between objects during an interactive naming task, provides clear nonverbal confirmation of referent selection, and supports a pilot crowdsourced evaluation that queries observer confidence—preliminary results suggest improved confidence but remain exploratory.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:20Dynamic Gaze Behavior
- 0:41Interactive Object Learning
- 1:38Prior Approaches
- 2:34Our Approach
- 4:20Integration
- 5:11Demo
- 6:12Demonstration
- 7:55Next Steps
related talks
14mTowards Unobtrusive Physical AI: Augmenting Everyday Objects with Intelligence and Robotic Moveme...
08mA Simple Nod of the Head: The Effect of Minimal Robot Movements on Children's Perception of a ...
21mTeaching Language to Deaf Infants with a Robot and a Virtual Human
18mRobotics: why now?
28mPreference Learning from Minimal Human Feedback for Interactive
15mSensing and Handling Engagement Dynamics in Human-Robot Interaction Involving Peripheral ...
15mRobot Hand Driven by Computer Vision and Reinforcement Learning
18mOn the History and Future of AI