wilamowski-b-m-irwin-j-d-industrial-communication-systems-2011

10-6 Industrial Communication Systems
preamble and stay awake to receive this message. The drawbacks of this approach are overemitting for
the preamble and overhearing at nodes not involved (addressed) in the communication as they need to
stay awake until the preamble ends to check the message. More energy efficient are X-MAC [BYAH03]
and CSMA-MPS [MB04], where multiple short messages, with the relevant node address, are repeated
instead of a single long preamble. This not only avoids the overhearing of nodes that are not involved, but
also allows interested nodes to indicate when they are awake and ready to receive. The downside of these
asynchronous approaches is that the energy problem is mainly shifted to the sender, who needs to send
the preamble. The length of the preamble depends on the check interval of the receiver and the energy
efficiency becomes a trade-off between overemitting at the sender and idle listening at the receiver.
Rendezvous-based approaches synchronize the wake-up periods of a group of nodes. The first
approach was the S-MAC protocol [YHW04], in which all nodes awake periodically for a fixed time to
exchange messages. IEEE 802.15.4 uses a comparable approach in its beacon-enabled mode. As in both
approaches, nodes not involved in communication also stay awake for the active time; this results in
overhearing and idle-listening again. T-MAC [DL03] improves this issue by setting uninvolved nodes
to sleep. The downside of all rendezvous-based approaches is that all nodes wake-up at least for the
synchronization, and the energy consumption depends strongly on the wake-up cycle, which is usually
constant. Especially in scenarios with a dynamic communication load, this leads to idle listening,
when no node has a message to transmit, and collisions, when multiple nodes intend to send messages.
The adaption of the wake-up cycles to the traffic load of the nodes is therefore one approach to reduce
energy-consumption while providing a high throughput [NPK05].
Rendezvous-based approaches are usually more energy efficient than asynchronous approaches. But,
application requirements should also be considered when deciding about communication protocols.
For example, if a temperature sensor samples a new value every minute, then the node should not have
to synchronize in a rendezvous-based MAC with a node sampling the illumination every 100.ms. In
this case, an asynchronous approach may be a better choice, but then the preamble for the temperature
sensor might be very long. Hence, network topology, communication protocols, and application design
need to be harmonized for ultralow-power communication.
10.4 application Layer Approaches
An appropriate application design is essential for ultralow-power communication. Again, the application
requirements of the node need to be considered to choose the most energy-aware design. Two general
design principles should be used: first, allow the node to sleep as often and as long as possible; and second,
process information locally, as the data processing consumes significantly less power than transmitting
data [RSS02]. Thus, the number of transmitted messages, the message size and the number of wake-ups
should be reduced, while still providing the required functional quality. For example, if the network is
used for monitoring applications with no hard real-time requirements, then gathering several samples
in the node and transmitting them in one compressed message can save much energy, since larger communication
intervals can be selected and the message overhead is proportionally smaller [MDG06].
If the network has a tree topology and only aggregated data (e.g., mean temperature) is needed, then
the information of individual sensors (temperature) can be combined and only the aggregated data can
be forwarded to the end user [KEW02].
If the network is used for monitoring or sensing purposes with hard real-time requirements, then the
node needs to wake-up regularly to sense changes in the environment. The common choice would be
periodic sampling as it is well covered by theory. This means the node wakes up in constant intervals for
sampling and transmitting each sample (Figure 10.6a). However, real-world signals usually change their
dynamics over time. The illumination is one extreme example as it can change during the day with moving
clouds and objects instantly, while nothing changes at night when it is dark. The frequent changes
during the day require a small sampling period, which is inefficient in the night with only small changes
and sampled values that are very similar.
© 2011 by Taylor and Francis Group, LLC