Deep Reasoning in Clarification Dialogues with Mobile Robots

Abstract

This paper reports our work on qualitative reasoning based clarification dialogues in human-robot interaction on spatial navigation. To interpret humans' route instructions, a qualitative spatial model is introduced which represents the robot's beliefs in the application domain. Based on the qualitative spatial model, three tool-supported reasoning strategies are discussed which enable the robot to generate dialogues with differing degrees of clarification if knowledge mismatches or under-specifications in route instructions are detected. The influence of the reasoning strategies on human-robot dialogues is evaluated with an empirical study.