A Service Recommendation Using Reinforcement Learning for Network-based Robots in Ubiquitous Computing Environments

Abstract

Ubiquitous robotic companion (URC ) is a new concept for the network-based robot platform which can enable to be following its master wherever or whenever he/she be in order to provide necessary services. The robot platforms in present normally interest in providing services through the direct interaction in responding to the user's demands. On the other hand, URC services are required to be provided by the means of recognizing the circumstances and taking a user's preference into account. In this paper, we propose a service recommendation scheme for URC robots. The proposed service recommendation, developed based on the reinforcement learning, can be used to provide personalized services by learning users' preferences or tasks through the interaction with users. Using simulation for rapid testing, we evaluate of the proposed scheme under a variety of user modeling types and discount factors.