The NARF Architecture for Generic Personal Context Recognition

Ubiquitous computing envisions applications that provide seamless and distraction-free support for everyday tasks. To achieve this goal, applications must be able to adapt to their environment and to the intents of their users. Thereby, they need to automate adaptation decisions to minimize the resulting distraction. As a consequence, it is necessary to acquire an understanding of the user’s current situation to ensure that the automation results in a desirable application behavior. The wide-spread use of personal mobile devices provides a promising technical basis to acquire this knowledge in a seamless manner. However, in order to account for the resource limitations of such devices, existing personal context recognition systems are typically highly specialized and cannot be adapted easily to different scenarios. In this paper, we motivate the need for an generic personal context recognition system and we derive the requirements on such systems. As an approach towards creating such a system, we describe NARF - our ongoing effort to create an adaptive context recognition framework. To assess the benefits and limitations of the architecture, we discuss an application that we have built using our current prototypical implementation of this framework.