|
The topic of this dissertation is the observation of physical world
events through a distributed world model. So the events of interest
occur in the world we live in. The basis for their observation is a
model of the relevant aspects of the physical world. These include
more static aspects like geometric models of stationary objects,
e.g., houses and streets, but also dynamic aspects, e.g., the
position of mobile users or the temperature.
With the proliferation of mobile computing devices like personal
digital assistants or mobile phones with significant computing and
communication capabilities, there is a trend to extend computer
support from the desktop to the physical world. As the focus of the
mobile user may be on other tasks, computer support should be
proactive, providing the user with information and services relevant
in his current situation. The observation of high-level physical
world events is an enabler for these new kinds of services.
Due to the size of the data, different characteristics of the data,
and a multitude of providers, the world model data needed for the
observation can be distributed over a number of servers. We present
a novel event service architecture that allows the observation of
complex high-level events through a distributed world model.
As the accuracy of the data is limited due to the characteristics of
both the underlying sensor data and the computer network, this has
to be taken into account. We propose a concept for specifying
physical world events together with a threshold probability above
which the event is considered to have occurred. We then show how
physical world events can be observed, calculating the occurrence
probability and comparing this to the specified threshold
probability.
Finally, we present an evaluation based on a prototype
implementation with a number of concrete events. The focus of the
evaluation is on both the performance and the quality of the
observation, showing the general feasibility of our approach.
|