High Level Definition of Event-Based Applications for Pervasive Systems - Advanced Design Approaches to Emerging Software Systems - page 140

start with an idea of the phenomenon that shall
be detected by the WSN. Obviously, the user has
to know the properties of the phenomenon to be
able to model it. This knowledge about properties
may be even very high level, e.g., “a fire is hot
and produces smoke”. An experts system or WSN
configuration assistant will be able to guide the
user with questions as simple as possible. Such
assistant is drafted in Figure 3. It will request the
properties of the phenomenon and based on that,
the boundaries of respective physical measure-
ments, which are typical for the phenomenon in
mind. Without knowing, the user already specified
primitive events. In case multiple measurements
jointly describe the phenomenon, mutually or alter-
natively, the user can be supported in combination
and arrangement of these measurements. It is easier
for users with no programming or mathematical
background to understand the semantics of words
like: together, mutually, alternatively or sequen-
tially, than to understand logical operations and
relations. Since the assistant uses those simpler
words in its questions, complex descriptions of
phenomena can be created by ingenuous users.
The intelligence to transform a together into an
AND or even a mutually into “this AND NOT
that, OR NOT this AND that” is embedded in
the assistant and concealed from the user. By just
these few questions the user is guided to formally
describe a phenomenon.
A complete event specification requires con-
straints that have to be defined by users as well.

Those constraints are, for instance, sampling in-
tervals and regions of collaboration. Also these
can be derived from the answers to more intuitive
questions, e.g., “After what time the phenomenon
has to be detected latest?” or “How long have
batteries to last?”. Without explicitly known to
the user, the EDT provides robustness through
collaboration of neighboring sensor nodes with
heterogeneous sensing capabilities. The distance
or relative area in which neighboring nodes shall
collaborate is a constraint that has to be set in the
event specification. As we learned from our
simulations and measurements, the appropriate
size of the collaboration region can be derived
from the expected expansion of the phenomenon
and the average distance of deployed sensor nodes.
The minimum size of the collaboration region
should be the mean distance between neighboring
sensor nodes, which is determined by the density
of the sensor network, and transmission technol-
ogy. The maximum size of the collaboration region
is the estimated size of the phenomenon. Both
constraints applied to these parameters result in
an appropriate region size. These parameters can
be queried from the user with simplified questions,
e.g., “What is the estimated diameter of the phe-
nomenon expected?” or “How large is the area in
which the WSN is deployed?” and “How many
sensor nodes the WSN consists of?”. Despite a
circular region will likely be sufficient for most
phenomena, different shapes can also be consid-
ered.

Figure 3. Guided derivation of machine processable event specification from vague human ideas.