OPPORTUNITIES FOR FUTURE DEVELOPMENT
One is tempted to state that, since the technology is only
now emerging from academic research into commercial application, all areas of
wireless sensor networks represent opportunities for future development.
However, some areas merit particular attention.
Opportunities for future work on the physical layer include the
addition of higher layers of the protocol stack to the simulations presented in
Chapter 3, so
that package error rate and message error rate may be simulated and the whole 2.4-GHz coexistence question may
be considered. If the duty cycle of a wireless sensor network node is 0.2
percent, can it be a significant interference source to a Bluetooth piconet?
Other services, such as IEEE 802.11b Wireless Local Area Networks (WLANs) and
microwave ovens, exist in the 2.4-GHz band; the performance of wireless sensor
networks in their presence is an interesting study, hinging first on the
definition of performance used and then on the type of application to which the
wireless sensor network is applied. Similarly, the performance of other services
in the presence of wireless sensor networks is an interesting subject.
The related problem of multiple access protocols in the
Industrial, Scientific, and Medical (ISM) bands also is waiting to be solved.
Methods to quickly determine the nature of signals present in the channel, not
just their existence, would greatly help the coexistence problem now facing 2.4
GHz, and certain to arrive at 5 GHz.
As semiconductor processes improve, one may turn to the 24- and
60-GHz ISM bands where, due to the reduced wavelength, interesting features such
as integrated antennas[2]
may be considered in wireless sensors. This would further reduce their size, and
perhaps create new applications for them, but the optimum design of such
antennas remains an open issue.
Turning to the mediation device protocol, at present, the design
assumes a simple random process to determine the MD period. The MD period could
be made dynamic, and a function of the number of node neighbors, the amount of
recent message traffic, or the amount of energy remaining in the node's power
source. The effect of this dynamic MD period on the node duty cycle, network
throughput, and message latency are open questions. The MD behavior could be
tied to network quality of service (QoS) provisions, and respond to a QoS field
in the message header or separate control messages.
One can envision that, if the offered message rate were low
enough, a lower-power algorithm could eliminate all regular MD activity in the
network altogether, and replace it with the "emergency mode" MD described in Chapter 4, Section 4.3.3. As
the offered message rate rose, however, the constant scanning of the
neighborhood before every message would result in power consumption higher than
the MD protocol. A protocol could then be devised that changed modes as the
traffic rate varied, to maintain optimum power consumption. This would include,
perhaps, operating in "emergency mode" but maintaining a record of neighbors'
beacon timing so that, for multiple messages sent in succession, the receive
period could be discarded for the follow-up messages.
The physical topology of the network needed for maximum MD
throughput is not known. What is the optimum number of neighbors, or children,
or cluster size? This is of special importance when the nondistributed form of the MD protocol is used, and
special-purpose MD nodes are placed in the network.
MD nodes (both distributed and dedicated) could be given a message
store-and-forward capability, instead of the simple node synchronization
function described in Chapter 4. This would be analogous to the more common use of
the telephone answering machine, wherein the caller does not suggest a time for
the return call, but actually leaves a message. How this would affect network
performance is an interesting question. It may increase message latency in some
cases by requiring two hops where nodes could have transferred the message
directly, but it could also improve the connectivity of the network, by linking
two nodes that otherwise could not hear each other.
Although it was designed for operation in wireless sensor networks
in which nodes are assumed to be static, the use of the dedicated MD protocol in
mobile networks presents some interesting possibilities. The plasma metaphor
could be used to describe a network of MD nodes with store-and-forward
capability in a network of highly mobile nodes. The heavy, positive ions of the
plasma (the MDs) are relatively stationary, while the light, negative electrons
(network nodes) move at a high velocity, storing messages with the nearest MD as
they pass by for later forwarding by the intended recipient when it is next in
range. This leads to the use of MDs as data fusion/aggregation devices, in which
redundant data is eliminated, meta-data is created, and only a minimum number of
messages are transmitted.
Further, the fault tolerance of any hierarchical tree is always a
question; in this case, it is not known if the dependence of the network on the
MD for communication link establishment affects the fault tolerance of the
network. It is possible that the failure of one or more MD could partition the
network.
The development of routing algorithms for wireless ad hoc networks
is perhaps the most active area of research related to wireless sensor networks;
many approaches, from those based on the behavior of ants to those based on the
hierarchical structures of military organizations, have been proposed. Does a
single algorithm exist that is flexible enough to meet the needs of both
wireless mice and military sensing applications, while simple enough to be
implemented in the most inexpensive applications?
Chapter 5 described a network with the "monocrystalline"
network association model (i.e., growth of the network starts with the DD). A
network designed after the "polycrystalline" network association model, in which
network growth starts with ordinary nodes, would make an interesting
comparison.
One may argue that the most
useful feature of wireless sensor networks is the low power consumption of their
nodes, which makes them practical for a variety of new uses. In this area,
perhaps the most intriguing areas of research are the development of true low
power mixed-signal CMOS processes (i.e., processes designed for low power
operation rather than maximum speed), and the development of practical energy
scavenging systems.
Although low-power CMOS processes have been in production for many
years, their use has, in general, been limited to specific low-leakage
applications, such as digital watches. Their performance, although adequate for
that task, is, in general, insufficient for mixed-signal designs at 900 MHz and
above. However, much of the improvement in high-performance CMOS has come from
the reduction of parasitic capacitance and the reduction of supply voltages, two
performance improvements that also reduce power consumption, and a true
low-power, mixed-signal CMOS process, perhaps on SOI, would be a boon to the
wireless sensor network market.
The development of a wireless sensor network node with a practical
energy scavenging system would certainly receive publicity similar to the wide
acclaim received by Reutter with the introduction of his perpetual clock in
1928.[3] In addition to
the favorable publicity, the development of a practical energy-scavenging system
would almost certainly enlarge the total available market for wireless sensor
networks by making new applications practical.
Finally, probably the greatest opportunity for future
development lies in the actual application of wireless sensor networks. The
opportunity for the development of creative uses for these small, inexpensive,
low-power, self-organizing communication systems appears limitless.
Sections
Archive
Syndication
