You are here

Robot Perception Errors and Human Resolution Strategies in Situated Human-Robot Dialogue


Niels Schutte, John Kelleher, Brian MacNamee

Publication Type: 
Refereed Original Article
Robots that interact with a human user through natural language in a spatial environment present a case of situated dialogue. The distinctive characteristic of a situated dialogue is that each participant has a speci c perceptual perspective on a shared spatio-temporal context. Consequently, participants in a situated dialogue can not only make references that are evoking (i.e., denoting entities in the interlocutors' conceptual knowledge) and anaphoric (i.e, denoting entities that have previously been mentioned in the dialogue), but can also make exophoric references (i.e., references denoting objects in the shared context of the dialogue). Therefore, in order to participate in a situated dialogue, robots must be able to perceive their environment and to communicate with the user about what they encounter in the world [18]. If the user's perception of the world and the robot's perception diverge (e.g. due to problems in the object recognition software used by the robot [25], or mismatches in the user's and the robot's understanding of spatial relations [4, 17]), misunderstandings may arise in the dialogue. In this paper, we investigate the e ect of perceptionbased errors on human-robot dialogue, and how misunderstandings that arise from such errors are resolved.
Digital Object Identifer (DOI): 
Publication Status: 
Date Accepted for Publication: 
Monday, 31 October, 2016
Publication Date: 
Journal Advanced Robotics
Pages 1-15
National University of Ireland, Dublin (UCD)
Open access repository: