Indoor location tracking based on wireless sensors such as Wi-Fi and Bluetooth has recently received increased public attention in the context of location-based services (e.g. location-based advertising) and indoor navigation. However, until now, no clear best practices for filters, algorithms and sensor fusion have emerged, and, in fact, different use cases may require different sets of techniques for optimal results. In this paper, we thus present a sensor fusion approach that can integrate results from heterogeneous tracking algorithms based on discrete, symbolic locations and a reference architecture based on a pipeline approach that offers the necessary flexibility to easily adapt location tracking applications or -services to both future discoveries in tracking techniques and varying use cases. We also report on a comprehensive evaluation of a tracking application we built using the named architecture and sensor fusion in a large (ca. 2600 square meters) open space office under day-to-day business conditions.

Activity theory has gained increasing popularity as a tool for the design of interactive systems. Its central idea is that user actions can only be fully understood when analyzed within the context of motive-driven activities. However, this context is currently only partially explicit in existing models; e.g. the well-known model by Engestrom promotes social aspects of context, but does not make the physical and technical environment --- relevant aspects for interactive systems design --- explicit. We present a model that incorporates the social, physical and technical context of activities. This model supports designers and developers in understanding users, their activities and the respective contexts, and allows deriving requirements for the design process directly from the model. We believe this will prove especially helpful when designing for novel devices and multi-user scenarios. We illustrate the practical use of our model with examples from two domains: composition of IT services and co-located collaborative modeling.