Long-term Human-robot Social Interactions Using Compositional Multimodal Agent Models at Monash University

We are excited to offer a fully funded PhD position at the Faculty of Engineering, Monash University (Australia). This project focuses on developing new algorithms to equip social robots with the social, cognitive, and communicative skills needed to autonomously engage in meaningful, long-term human-robot interactions.

Project overview: Social robots are designed to be competent partners that help people achieve positive outcomes in diverse areas of their lives including education, and physical and mental well-being. Despite their potential, existing social robots often fail to be adopted and accepted by the intended users due to their limited capabilities to sustain long-term interactions. In this project we propose to develop compositional vision-language models for social robots, enabling them to build the requisite knowledge and skills required to adequately interact with a user. These models will allow social robots to recognise, understand, and predict a user’s behaviour, reason about how the robot’s future actions might influence the user and the context in which the interaction takes place, and determine whether an error has occurred and how to recover from it.