Abstract Personal Assistive Robotics

How can robots cooperate and assist humans in achieving their tasks? While some technical aspects might differ between specific tasks, there are fundamental challenges underlying all such endeavours: robots need to understand the current state and intention of their users through multimodal perception algorithms, and personalise assistance to the users’ particular (and evolving) needs.  In this talk I will present our research towards designing and building cognitive architectures that enable robots to build multiscale models of their users, and adapt their behaviour to maximise assistance effectiveness over extended periods of interaction. I will argue that such personalisation, and the explainability of the learned user models, can help in developing effective and trustworthy robot assistants. I will illustrate our approach with examples from assisting (children and adults) with activities of daily living, for example dressing and mobility, as well as educational tasks.