Leveraging Human Interaction to Learn and Reason over Multimodal Object Affordances
Abstract
The performer proposes to develop a domain-independent robotics framework that leverages semantic reasoning and situated interaction with a user and the environment in order to enable robust task execution in complex, novel environments. Specifically, their approach leverages three data sources - semantic networks, human interaction and exploration of the environment to 1) mine semantic language resources to construct an affordance knowledge base specific to the objects and locations in the robot s domain, 2) ground the knowledge base by modeling visual, tactile and auditory attributes and usability a affordances of objects in the environment through situated interaction, and 3) enable the robot to adapt to changes in its environment through inference over the resulting knowledge base.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 30, 2016
- Source ID
- N000141612835
Entities
People
- Sonia Chernova
Organizations
- Georgia Tech Research Corporation
- Office of Naval Research
- United States Navy