Answer me this: Constructing Disambiguation Queries for Explanation Generation in Robotics
Tiago Mota, Mohan Sridharan · 2021
Our architecture seeks to enable robots collaborating with humans to describe their decisions and evolution of beliefs. To achieve the desired transparency in integrated robot systems that support knowledge-based reasoning and data-driven learning, we build on a baseline system that supports non-monotonic logical reasoning with incomplete commonsense domain knowledge, data-driven learning from a limited set of examples, and inductive learning of previously unknown axioms governing domain dynamics. In the context of a simulated robot providing on-demand, relational descriptions as explanations of its decisions and beliefs, we introduce an interactive system that automatically traces beliefs, and addresses ambiguity in the human queries by constructing and posing suitable disambiguation queries. We present results of evaluation in scene understanding and planning tasks to demonstrate our architecture’s abilities.