Robot-wearable conversation hand-off for AI navigation assistant

Abstract

Indoor navigation systems today are deployed as static devices and as mobile applications. Promising virtual helpers based on artificial intelligence to assist users with various tasks, AI assistants and specifically voice-based AI assistants are becoming more capable and available. As a result, we begin to leverage their strengths in a wide variety of contexts.

In this thesis, I develop a conversational, voice-based AI agent to explore how AI assistants could improve the way we navigate indoors. Three types of embodiments are evaluated in comparison for this assistant. A traditional social robot embodiment is compared with a wearable embodiment and with a novel conversation hand-off system. A user study with 12 participants was conducted, where I collected task performance, perceived utility and user experience data.

The data show that the social robot embodiment requires higher mental effort compared to the other two methods, and that participants are divided about the potential value of the conversation hand-off. The results set a direction for further development of CA for indoor navigation and for the proposed conversation hand-off method.