More Irish Countryside Adventures!
SW Architecture Workshop Summary
Wearable and Pervasive Computing
We Don't Get No Respect! (Panel)
Window on the World
Issue 2
Workshop 7: Wearable and Pervasive Computing
By Michael Karasick
While there is little agreement on the details of software architectures for wearable and pervasive computing, several broad themes emerged over the course of the workshop:
(1) Multi-layer architectures are applied universally to abstract away the particulars of wearable devices, sensors, and pervasive computational elements.
(2) Layers exist to erase minor differences among like sensors, to transition seamlessly between distinctly different sources for an abstract datum (for example, GPS for location information outdoors, and infrared ranging for location detection indoors), and to provide adaptation or transcoding services for diverse devices.
(3) Higher levels of these same architectures are required to provide planning, inference, learning, and constraint management services for driving applications.
(4) The separation of concerns dominates the structure of these architectures, with individual layers targeted towards resource management, energy conservation, context- and user-interface adaptation.
There was uniform agreement that events are the fundamental connective glue for the individual system components. However, concerns were raised regarding the distribution of events, their scalability, their transparency, and their semantics. It was suggested that declarative policies are an appropriate mechanism for managing this complexity. Example implementations were dominated by flexible component models and proxy architectures whose primary function is to allow partitioning of applications between devices, end-of-network servers, and back-end servers.
Sensors were a recurring obsession. Architects are struggling with sensor semantics, aggregation, fusion, representation, and even basics such as a sensor taxonomy. This is in sharp contrast to our ability to abstract, manage, and deliver content effectively to dissimilar client devices. Finally, it is clear that the systems require substantial implementation and that results are impossible to obtain until the infrastructure has stabilized so that systems and components can shared among researchers.
Designer
and Webmistress
Caroline Kehoe