UKRI AI CDT in Dependable and Deployable AI for Robotics (CDT-D2AIR)

Ruchira Ray

Project
Active Vision for Robot Learning

Summary
I am broadly interested in the intersection of active perception and learning from demonstration. My research focuses on how robots can learn more effectively by actively acquiring information and closing the loop between perception and interaction. In my research, I currently aim to investigate how active perception can improve learning from demonstration by enabling robots to dynamically select viewpoints while observing a teacher. This work aims to improve learning under partial observability by reducing ambiguity, handling occlusions, and focusing on task-relevant information.