AdAPT -- A Dynamic Approach for Activity Prediction and Tracking for Ambient Intelligence

Abstract

With recent advancements in supporting fields like embedded systems and Ambient Assisted Living (AAL), intelligent environments are becoming reality. However, instrumenting an environment with a set of sensors and actuators and applying some automation rules alone doesn't make the environment intelligent. Learning and adapting to user behaviors and gaining some basic knowledge about the underlying intention is an essential feature of an intelligent system. Here, we introduce AdAPT, which is an incremental approach for recognizing, predicting and tracking Activities of Daily Living (ADLs) within a smart home infrastructure. Our approach does not make any predefined assumptions about typical activity models but tries to learn and adapt to the user's actual behavior continuously. We focus on designing suitable interaction concepts to support an optimal, continuous and unobtrusive adaption to the user. In this paper, we introduce the AdAPT project, highlight relevant research questions and provide a first description of the proposed system design.