Applications and Future
We have covered many important topics on designing and deploy-
ing large-scale sensor networks in the previous chapters. Our pri-
mary focus has been the information processing aspects of sensor
networks—namely, how sensor information is acquired, represented,
processed, transmitted, aggregated, and accessed. This chapter will
summarize the salient points in these discussions, describe emerging
sensor network applications, and outline future research directions.
8.1 A Summary of the Book
We have advocated a holistic approach to the design of sensor
networks, including optimizing across physical layers, networking,
embedded OS, node services, group management, collaborative pro-
cessing, and application-specific needs. Figure 8.1 shows one possible
organization of a sensor network stack. While the exact layers and
interactions are still debatable and depend on the particular sys-
tem requirements and environmental constraints, a sensor network
stack must at least support information processing across multiple
nodes in a resource-aware manner. Hence, the central theme of what
we have presented so far can be summarized as developing scal-
able algorithms and software architectures to support distributed
information processing applications on resource-constrained sensor
To describe the key elements of a typical sensor network system,
we introduced distributed object localization and tracking as a
292 Chapter 8 Applications and Future Directions
APIs and programming models
Query processing/data management
Information fusion
Collaborative group management
Storage, time synch, location services
Signal proc. Signal proc.Signal proc.
Processor ProcessorProcessor
Sensors SensorsSensors
Figure 8.1 The organization of a sensor network stack.
canonical problem (Chapter 2). We studied a number of information
estimation and fusion techniques that are well suited for distributed
signal processing and discussed the information processing and com-
munication requirements for the tracking problem. To meet these
requirements, we presented techniques such as geographic and
attribute-based routing for sensor networking (Chapter 3). The key
idea was to name and route data based on physical attributes such as
location rather than logical (IP type) node addresses. We investigated
a number of techniques for establishing a common time base across
8.1 A Summary of the Book 293
nodes and for nodes to discover their locations (Chapter 4); these are
important services that enable networks to initialize themselves and
carry out collaborative sensing and processing tasks.
Sensor tasking and control is an important topic for resource-
constrained sensor networks. We introduced a number of
information-driven approaches that optimize for information gain
and resource utilization by solving a distributed constrained opti-
mization problem (Chapter 5). Intrinsic to these techniques is the
notion of collaborative group management that aids the forma-
tion of dynamic, ad hoc groups of sensors to support distributed
processing tasks. Treating a sensor network as a distributed database,
we studied the problem of storing, locating, and accessing data in
a distributed, power-aware setting (Chapter 6). We discussed the
topics of in-network aggregation, data-centric storage, range search,
and multiresolution summarization. Finally, we studied the problem
of system architectures and programming models for sensor net-
work applications (Chapter 7). Because of the cross-node interactions
and event-driven nature of many sensing applications, appropri-
ate software architectures have to effectively manage concurrency
and resource allocation, while presenting to application developers
a model of programming that is closer to application semantics than
to lower-level system programming.
An important lesson we can draw from these studies is the
unavoidable interleaving of information processing and software
architecture in sensor network systems, and the need to co-design
and co-optimize these two architectures simultaneously. For exam-
ple, in the IDSQ algorithm we discussed (Chapter 5), the information
flow from node to node, as dictated by the Bayesian probabilis-
tic tracker, requires the corresponding networking protocols, group
management, and software frameworks to support this style of infor-
mation processing and communication. As we gain more experience
with a broad range of sensing applications, we can expect to discover
other important patterns in the interactions between the informa-
tion processing and software architectures, and design algorithms
and tools to exploit these patterns.

Get Wireless Sensor Networks now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.