Maximizing users comfort levels through user preference estimation in public smartspaces

In recent years, the ubiquitous computing system attracts people's attention as the system to provide useful services (e.g., automatic temperature control) without explicit operations by users. There are many existing methods for controlling appliances according to the user's preference by describing each user's preference and rules. However, these methods cannot be applied to public spaces where many general users with different preferences exist. In this paper, we propose an architecture and a method for controlling devices that affect the users comfort level (e.g., air conditioner) in public smartspaces. Our goal is to maximize the comfort level of users with various preferences by appropriately controlling devices. Furthermore, to efficiently collect user preferences for the large context domain, we propose a method for estimating user's comfort level for an unknown context from the already known user's comfort level for some contexts and the distance to those contexts. To evaluate the proposed estimation method, we conducted the questionnaire to measure the user's comfort levels for various contexts, and evaluated the accuracy of the proposed estimation method by comparing the measured sample with the estimated one. As a result, our method estimated user's comfort level in error within 1 among 4 comfort levels.