Acquiring Grounded Representations of Words with Situated Interactive Instruction
This addresses the challenge of building more capable and adaptable AI agents for robotics, though it appears incremental as it builds on existing interactive learning methods.
The paper tackles the problem of acquiring grounded word representations by enabling a robotic agent to learn from interactive human instruction, resulting in efficient learning of perceptual, semantic, and procedural knowledge.
We present an approach for acquiring grounded representations of words from mixed-initiative, situated interactions with a human instructor. The work focuses on the acquisition of diverse types of knowledge including perceptual, semantic, and procedural knowledge along with learning grounded meanings. Interactive learning allows the agent to control its learning by requesting instructions about unknown concepts, making learning efficient. Our approach has been instantiated in Soar and has been evaluated on a table-top robotic arm capable of manipulating small objects.