Task-based prediction of interaction patterns for ambient intelligence environments

Abstract

In this paper we introduce a monitoring system to support the user executing tasks in an ambient intelligence environment. In contrast with traditional environments, the goal of the user can not always be defined beforehand, but is determined while the user interacts with the environment. The monitor observes the user's activities and learns to correlate a set of user actions with a goal. The system maps activities to a task model and reuses these models to take appropriate actions in later similar user actions that are observed.