High Level Definition of Event-Based Applications for Pervasive Systems - Advanced Design Approaches to Emerging Software Systems - page 170

use low cost sensor nodes with limited resources,
which feature a high fault probability. WSNs
are subject to sudden changes in operational
conditions, varying deployments and hazardous
environments that again increase the fault prob-
ability. Moreover, strict energy constraints on used
devices require fault tolerant methods to achieve
high cost-efficiency.
This chapter identified missing features for
convenient high-level application design for
WSN configuration. Therefore, we introduced a
user-centric design flow of pervasive applications.
It decouples the processes of application design
and WSN configuration. Application design is
advanced to a level that allows users to specify the
“things-in-mind to be sensed” without regarding
WSN properties. From our simulations we have
learned how to deduce technical details like the
collaboration region or the leasing time without
explicitly asking the user for. Finally, the user
is enabled to “configure” a WSN by answering
straightforward questions about the phenomenon
to be detected. Based on automatic generation of
event specifications using the ESL, the EDTs allow
to autonomously configure a WSN by submitting
very compact binary event specifications. Reliable
and robust execution of applications in pervasive
WSNs further requires to cope with expected and
unexpected heterogeneity and sudden failures.
Therefore, we introduced objectives for reliable
event-based applications in WSNs in terms of
design criteria. These are Robustness , Autonomy ,
Transparency , Energy efficiency and Convenience .
Existing solutions mostly provide Robustness and
Transparency but disregard sufficient Energy ef-
ficiency , Autonomy and Convenience . It has further
been shown that existing solutions lack of means
to achieve an acceptable cost-efficiency.
Our introduced user-centric design and con-
figuration of reliable event-based applications in
WSNs can actually remedy these shortcomings.
It tackles all design criteria and features cost-
efficient robustness and a proper usability. So it
combines a flexible Event Specification Language

with a self-adapting event-based detection scheme.
The ESL provides ease of use for application pro-
gramming allowing the user to ignore low-level
details of the sensor network and to concentrate
on a high abstraction level. Namely this is the
phenomenon itself and its related constraints.
To cope with the fault probability in WSNs,
communication-efficient means for collaborative
detection have been introduced and proven to be
functional. In detail, the following contributions
are made:
High abstraction for user-centric application
design. The ESL hides low level details of WSNs
to focus on pure phenomenon definition, which
allows automatic configuration of event detection
in WSNs. The ESL enables to combine sensing
features defining the complex phenomena to be
sensed. Further, it enhances an event specifica-
tion by assignment of customized application
requirements regarding the spatial and temporal
expansion and parameters for collaboration.
Fine-tuning of the collaboration procedure by
determining a proper collaboration region and suit-
able time limits is supported by the ESL. Finally,
the event specification generator transparently
processes and adapts the user-defined “things to
be detected” to the target sensor platform and its
possible heterogeneity. Thereby, it generates de-
ployable versions of these “things”, called binary
event specifications.
The ESL addresses the following design
criteria: Transparency , Energy efficiency , Con-
venience.
A novel decentralized mechanism to au-
tonomously set up event-based detection and
in-network processing on sensor nodes, called
EDT. Binary event specifications are deployed
on the sensor nodes as EDTs, which are directly
generated on the nodes by a tiny GFSM requir-
ing eight states only. An EDT enables the sensor
nodes to partition event specifications according
to their own resources into local and remote parts
by pruning. Local parts can be evaluated by the
node itself, whereas values of remote parts must