Image Processing Reference
In-Depth Information
11.1.1.4 Tight Integration
Real WSN deployments are tightly integrated into the environment rendering them extremely sus-
ceptible to the effects of harsh and considerably varying environmental conditions. his integration
is a commonly known cause of faults. Haneveld [] reports unanticipated obstructions: The leaves
of the plants present on the deployment site changed transmission characteristics when there was
rain. He observed a correlation between packet reception ratio, temperature, and humidity. In
the initial Great Duck Island deployment Szewzcyk et al. [] discuss node failures and attributes
a number of them to humidity and broken seals on the packaging. The impact of environmen-
tal conditions may be deeply rooted as in the case of the heavily used TI MSP []. It is a
microcontroller directly designed for ultralow power system design. Power mode changes take
effect immediately. Nevertheless, even such a targeted device needs to be handled with care for
long-term outdoor deployments: One example is the negative temperature coefficient of the dig-
itally controlled oscillator (DCO) when the DCO is disabled for long low-power sleep states. In
outdoor deployments, e.g., relying on scavenging energy from the sun, a system may be put to
sleep for prolonged periods of time during the night. Such a long-term sleep period can result
in a delayed wakeup of the DCO as the device is impacted by a variation in temperature. The
effect starts to severely show with sleep periods of tens of minutes and a temperature variation of
tens of degrees, which are common characteristics in environmental monitoring. This and similar
efectsaredocumentedintheineprintofthemanufacturerdocumentationbutotenoverlookedin
practice.
11.1.2 A Case for Coordinated Testing and Validation
The design of WSNs is hard: Failed and unsuccessful deployments [] underline the need for (a)
better design tools and methodologies and (b) for an increased focus on the validation of WSNs.
SomeofthechallengesofWSNsare
•
Large state space of the sensor nodes due to interrupt-driven software with preemptive
scheduling
•
Vast distributed state space of highly concurrent sensor nodes
•
Dynamically varying spatial locality due to unreliable, broadcast medium
•
Visibility of the system is limited and expensive, since data traffic is low and energy scarce
•
In-depth optimization of resource consumption
•
Tight integration into environment increases susceptibility and dependencies on envi-
ronment
•
Integration of issues and challenges from differing fields (wireless, embedded, mechani-
cal, and concurrent) into an integrated design methodology
Therefore, the testing of WSNs must encompass various aspects of the system under development.
Tests and testing tools need to consider the intricacies of the embedded devices combined with the
complexities of testing a distributed system. In addition, the unreliable and heavily bandwidth con-
strained radio communication aggravates these issues. hus validation and testing are considerably
difficult and complex.
The following section presents WSN validation using testing techniques and test platforms specif-
ically designed for the sensor network case. A classification of test platforms is provided to illustrate
the applicability of a platform for certain classes of tests. A prominent and widely used test platform
is a WSN testbed, which is presented in Section . using the deployment-support network (DSN)
as an example. Finally, a test architecture is presented which integrates established methodologies to
address the common validation issues in WSN development.
