Robotic companions have the potential to revolutionize daily life. The integration of artificial intelligence (AI), Large Language Models (LLMs), and robotics into home settings presents an opportunity to improve safety, efficiency, and quality of life for individuals. However, technological progress on robotic companions are accompanied by distinct challenges, two of these are the presence of representative household simulation environments, and the ability to interpret and respond to user commands in real time, which is often complicated by the ambiguity inherent in human language.

In this proposal, we describe contributions along three aspects of making robots better companions: creating simulation platforms with plausible human agents, user communication and resolving ambiguities, and finding and responding to household dangers. By approaching these areas of robotics, we aim to work towards new personal robots who actively improve their users lives.