Exploring Shared Mental Models in Household Human-Robot Teams

In human-human teams, we often infer the situation awareness of our teammates to inform our planning and decision-making. What if we applied this to human-AI teams? This project explores how autonomous systems can estimate the belief states of human teammates. We deploy a robot agent to a collaborative household cooking domain, where the agent constructs and maintains a model of what a co-located human is aware of. However, current methods assume that users have a perfect perception of their environments. Through comparison with user study situation awareness responses, this project develops methods to more closely align the constructed model to the capabilities of users.