A corpus-based analysis of route instructions in human-robot interaction

Abstract

This paper investigates how users employ spatial descriptions to navigate a speech-enabled robot. We created a simulated environment in which users gave route instructions in a dialogic real-time interaction with a robot, which was operated by naïve participants. The ability of robot monitoring was also manipulated in two experimental conditions. The results provide evidence that the content of the instructions and strategies of the users vary depending on the conditions and demands of the interaction. As expected, the route instructions frequently were underspecified and arbitrary. The findings of this study elucidate the complexity in interpreting spatial language in HRI. However, they also point to the need for endowing mobile robots with richer dialogue resources to compensate for the uncertainties arising from language as well as the environment.