AlTair

Social and domestic robots are aimed at providing assistance and companionship in the daily life of elderly users. In this context, the robot's ability to automatically adapt to the user, user-adaptiveness, can be a deciding factor in the system's success. This work presents POMDP, a POMDP based decision-making mechanism able to learn and adapt to a user, applied to a realistic scenario. Specifically, this work introduces the following key innovative factors: A novel state-based reward formulation; A novel learning mechanism and execution loop.