About‎ > ‎


The idea of context recognition itself is not new. Researchers have successfully built systems that reliably detect various characteristics. Unfortunately, the existing solutions usually focus on simple characteristics or they only work in a narrowly defined setting. To support the recognition of complex characteristics in a broad range of scenarios, the NARF project is tackling the following research challenges:

  • Practicable Recognition: To support realistic future application scenarios, it is necessary to enhance existing and to develop new methods for context recognition that work well under real world conditions with reasonable costs for end users and minimal effort for calibration. For more information about practicalbe recognition check out our papers on indoor localization.
  • Dynamic Composition: To support the heterogeneity of user task envisioned by ubiquitous computing, it is necessary to dynamically compose context recognition methods that have been developed in isolation. Thereby, it is necessary to detect and avoid duplicate computations in order to safe resources. For more information about dynamic composition check out our work on the component system.
  • Privacy-preserving Collaboration: To support the ad hoc collaboration of devices, it is necessary to share potentially sensitive context information. To protect the privacy of users, it is necessary to design new concepts and mechanisms that limit the sharing of information without manual intervention in a desirable manner.
  • Automatic Personalization: To support different types of users, it is necessary to personalize the context recognition methods such that they meet the habits and preferences of the user. This requires the development of new tools and algorithms that optimize existing and even synthesize new ways of recognizing a particular context.

Continue reading about the basic architecture ...