Introduction to Optimization Goals and Figures of Merit

Optimizing wireless sensor networks (WSNs) requires addressing complex questions, such as
how to improve network performance, compare solutions, and support specific applications. A
major challenge is translating broad, imprecise goals into measurable metrics or figures of merit.
Given the diverse range of applications, a universal solution is impractical. However, focusing
on specific optimization aspects like QoS, energy efficiency, scalability, and robustness can help
tailor solutions to application-specific needs.

 Quality of Service (QoS)

Quality of Service (QoS) in WSNs differs from conventional networks, as these networks
primarily focus on transferring bits efficiently rather than delivering traditional communication
services. QoS attributes are categorized into low-level (e.g., bandwidth, jitter) and high-level
(e.g., event detection accuracy, subjective user experiences). In WSNs, high-level attributes are
often more relevant. Key QoS metrics include the probability of detecting and reporting events,
classification accuracy for detected events, detection and reporting delays, and the precision of
tracking and approximation tasks. These metrics vary based on application requirements,
emphasizing the need for tailored QoS solutions.

 Energy Efficiency

Energy efficiency is a critical goal for WSNs due to the limited energy resources available in
these networks. Various metrics evaluate energy usage, such as energy per correctly received bit,
energy per reported unique event, and delay-energy trade-offs, which prioritize urgent events.
Network lifetime is another vital metric, measured by time to first node death, network half-life,
or time to loss of coverage. These metrics provide insights into the network's longevity and
operational efficiency. However, evaluating energy efficiency requires assumptions about energy
consumption, network load, and environmental factors, making it a challenging but essential
aspect of WSN optimization.

 Scalability

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 1

Scalability refers to the ability of a WSN to maintain performance irrespective of network size,
an essential requirement given the potential for thousands of nodes in these systems. Achieving
scalability is challenging due to resource constraints and the need to avoid globally consistent
states, such as large routing tables. Protocol designs must balance the trade-off between
scalability and performance, particularly in smaller networks, which might allow for more
efficient solutions. Research in this area focuses on developing architectures and protocols that
support high scalability while accommodating application-specific requirements.

 Robustness

Robustness in WSNs ensures network functionality despite node failures, energy depletion, or
environmental disruptions. These networks should compensate for failures by finding alternative
routes or mechanisms to maintain operations. Robustness depends on failure models for nodes
and communication links, making precise evaluation challenging. Effective robustness strategies
focus on ensuring that limited node failures or changes in the environment do not significantly
impact network performance, thereby enhancing reliability and resilience in diverse applications.

Design Principles for Wireless Sensor Networks (WSNs)

When designing and optimizing wireless sensor networks (WSNs), key goals such as Quality of
Service (QoS) support, energy efficiency, and scalability are essential. However, these goals
alone do not provide a clear blueprint for structuring the network. To achieve these objectives
effectively, several fundamental design principles have emerged over time. These principles aim
to guide the development of networking protocols that support these goals while addressing the
specific requirements of each application. While these principles offer valuable insights, it is
crucial to consider the unique needs of a given application because blindly following these
principles may sometimes lead to suboptimal solutions.

 Distributed Organization

Distributed organization is fundamental to achieving scalability and robustness in WSNs. Unlike

centralized systems that rely on a single controlling entity, distributed organization ensures that
nodes collectively manage tasks such as medium access and routing, thus avoiding single points

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 2

of failure and adapting better to the limited communication range of WSN nodes. However, fully
distributed systems may suffer from inefficiencies compared to centralized ones. A hybrid
approach, where centralized principles are applied dynamically within localized regions through
elected nodes, can combine the strengths of both systems. This localized hierarchy must remain
dynamic to prevent resource depletion and maintain robustness, requiring continual re-election of
central nodes based on specific criteria. This dynamic self-organization process ensures efficient
network operation and adaptability.

 In-Network Processing

In-network processing leverages the computational capabilities of WSN nodes to process data
locally rather than transmitting raw data to a sink, significantly reducing energy consumption.
Techniques like aggregation, where intermediate nodes combine sensor readings (e.g.,
calculating averages), exemplify this principle. Another approach is distributed source coding,
exploiting spatial and temporal correlations in sensor data to compress it efficiently. Advanced
methods such as distributed signal processing allow nodes to collaboratively perform complex
computations, like Fast Fourier Transforms, reducing the amount of transmitted data.
Furthermore, mobile code and agent-based networking enable compact program codes to
traverse nodes, collecting data and making routing or fusion decisions. These methods enhance
energy efficiency, scalability, and network performance, making in-network processing a
cornerstone of WSN design.

 Adaptive Fidelity and Accuracy

WSNs must balance energy consumption with the fidelity and accuracy of the data they provide.
By adapting to the network's energy status, applications can determine the minimum acceptable
accuracy level for their needs. For instance, event detection applications can prioritize energy
efficiency by reducing the number of participating nodes or accepting less precise results. The
network should provide feedback on its operational conditions—such as energy availability or
node failures—to allow dynamic adjustments in application requirements. This adaptive
approach extends the principle of energy-aware computation from individual nodes to the entire

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 3

network, ensuring that the WSN continues to function optimally even under varying conditions,
a challenge that remains a significant research focus.

 Data Centricity

Data-centricity shifts the focus from addressing specific nodes to addressing the data itself. This
paradigm aligns with the application goals of WSNs, where the identity of the node reporting an
event is often less critical than the event itself. For instance, in an elephant-tracking scenario, the
application can specify interest in "elephant presence" rather than addressing specific nodes. This
shift enables more efficient communication, supporting in-network processing techniques like
data aggregation and fusion. Various implementation approaches include overlay networks with
distributed hash tables, which allow efficient data lookup, and publish/subscribe systems,
which enable nodes to request or provide data dynamically based on content rather than
identities. Additionally, treating WSNs as dynamic databases allows applications to query the
network for real-time or historical data, leveraging a data-centric structure for scalable, energy-
efficient operations.

 Exploit Location Information

Location information can significantly enhance the design and operation of communication
protocols in wireless sensor networks (WSNs). Since many applications depend on knowing the
location of an event, mechanisms to determine the location of sensor nodes and observed events
are essential. These mechanisms are detailed in Chapter 9. Once location data is available, it can
simplify protocol design and improve energy efficiency by enabling location-aware routing,
clustering, and data aggregation. Numerous protocols discussed in Part II illustrate how location
information optimizes energy usage and network performance.

 Exploit Activity Patterns

Activity patterns in WSNs differ significantly from traditional networks. Although the average
data rate over time may be low, rare events can trigger bursts of activity. This phenomenon,
known as the "event shower effect," occurs when an event is detected by multiple sensors
simultaneously, leading to a surge in network traffic. Protocols should be designed to handle

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 4

such bursts effectively, transitioning seamlessly between periods of inactivity and high activity.
This adaptability ensures efficient energy use while maintaining network responsiveness.

 Exploit Heterogeneity

Heterogeneity in WSNs can be leveraged to enhance network efficiency and reliability.

Heterogeneity may arise from:

1. Construction: Some nodes are equipped with larger batteries, higher communication
ranges, or greater processing power.
2. Evolution: Identical nodes start with equal resources but become heterogeneous over
time due to varying levels of activity or energy consumption.

Designing protocols that account for heterogeneity can optimize resource usage and extend the
network's lifespan. High-capability nodes can take on more demanding tasks like data
aggregation, routing, or acting as cluster heads, thereby distributing the workload more
effectively and reducing strain on less capable nodes.

Service Interfaces of Wireless Sensor Networks (WSNs)

 Structuring Application/Protocol Stack Interfaces

A key approach to structuring wireless sensor network (WSN) service interfaces is to treat the
application as just another component in the system. This is in line with the component-based
design discussed in Section 2.3. In such a model, applications can interact directly with other
components using predefined interfaces, such as the command/event structure used by TinyOS.
This means an application could consist of several components, each of which may interact with
others using the specified interfaces. This approach makes it easier to design flexible and
modular applications in a WSN.

 Expressibility Requirements for WSN Service Interfaces

For WSN service interfaces to be effective, they must provide certain functionalities to support
the various operational patterns typical in sensor networks. Key functionalities include:

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 5

 1. Support for Simple Request/Response Interactions:
o A service interface should enable simple request/response interactions, such as
retrieving data from a sensor or setting parameters on a node. These interactions
are synchronous, meaning the result (or acknowledgment) is expected
immediately.
o Additionally, the interface should support periodic interactions, where responses
may need to be provided on a regular basis, such as in applications requiring
continuous or time-based sensor data collection.
2. Support for Asynchronous Event Notifications:
o The service interface should also support asynchronous event notifications,
where a requesting node is notified about certain events without a fixed
relationship between when the request is made and when the information is
provided.
o These notifications could involve notifying a node if a certain condition or event
becomes true, such as a sensor detecting a threshold value.
o An important feature for such interactions is the ability to cancel the request for
information if it is no longer needed.
o Additionally, after an event occurs, the interface could allow for continued
actions, such as periodic reporting of data from sensors after an event.
3. Addressing Mechanisms:
o The service interface should allow the definition of communication peers for both
synchronous and asynchronous interactions. This can be achieved in different
ways, such as explicitly enumerating the peers or using a data-centric approach.
o One possibility is using a publish/subscribe model, where nodes are implicitly
grouped based on certain criteria, such as:
 Location-based groupings (e.g., nodes within a specific region of space).
 Observed value-based groupings (e.g., nodes reporting data that meets
certain criteria, like a temperature above a certain threshold).
o The interface should support operations like intersection, union, or difference
between such groups.
4. In-Network Processing:

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 6

 o The service interface should enable access to in-network processing functionality.
For example, when reading data from a group of nodes, the service should specify
if any in-network processing, such as data fusion or aggregation, is required.
o It should also allow applications to introduce custom processing logic into the
network. This could include defining new aggregation functions or deploying
mobile agents to execute specific tasks in the network.
o In some cases, in-network processing can be used to detect complex events,
which require data exchanges between multiple sensors to detect events that no
single sensor can detect on its own.
5. Accuracy and Energy Considerations:
o The service interface should allow the application to specify the accuracy of the
results required from a query. This could involve setting bounds on how many
nodes should contribute to the result or specifying how much compression should
be applied to the data.
o Along with accuracy, the interface must support the specification of energy
constraints to determine how much energy is acceptable to spend when
producing a particular piece of data.
6. Timeliness Requirements:
o The service interface should support timeliness constraints regarding when data
should be delivered. For instance, it may be acceptable to receive data quickly at
the cost of higher energy consumption (e.g., waking up nodes earlier or
transmitting data more frequently).
o Alternatively, the network could prioritize lower energy consumption by
delivering data more slowly, such as piggybacking information on other ongoing
data transmissions to conserve energy.

Gateway Concepts

 The Need for Gateways

A wireless sensor network (WSN) that operates independently is not practical for many real-
world applications. In most cases, a sensor network needs to communicate with other

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 7

information devices, such as mobile devices (e.g., PDAs) or remote users, to exchange data and
interact with the broader network, including the Internet. For example, a user might want to
access the data from temperature sensors in their home while traveling and connected to the
Internet via a wireless network.

To achieve this, the WSN needs a way to communicate with devices or gateways that bridge the
WSN with the broader Internet. This could involve a device like a mobile phone or a gateway
that offers an Internet connection. At the physical, MAC (Medium Access Control), and link
layers, this connection is relatively straightforward, especially if mobile devices or gateways use
similar communication technologies as the WSN, such as IEEE 802.11. However, a challenge
arises in the logical design of the gateway.

A simple approach might be to treat the gateway as a router between the Internet and the WSN,
potentially using Internet protocols within the WSN. However, this would be inefficient, as
WSNs generally require highly optimized, specific protocols. As a result, a more practical
solution is to design the gateway as an application-level gateway, where it makes decisions
based on application-level information. The complexity of the gateway's design depends on
where the communication originates.

 WSN to Internet Communication

When a communication originates within the WSN and needs to reach the Internet (for example,
when a sensor node wants to send an alarm message to an Internet host), the first problem is
discovering the gateway within the WSN. This requires solving a routing problem, similar to the
issues encountered in ad hoc networks, where the network needs to identify the route to the
gateway offering the required service. Service discovery mechanisms must be integrated with the
routing process.

Another challenge is how to choose among multiple gateways if more than one is available,
especially if not all Internet hosts are reachable through each gateway. To manage this, one
approach could be to build an IP overlay network over the WSN to handle this issue more
effectively.

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 8

A key issue is how to map semantic requests (like "Alert Alice") to concrete Internet addresses,
such as an IP address. Although the sensor node may not need to process IP protocols, it still
needs to embed the necessary information (such as the destination IP address and port) in its
packets. The gateway must then extract this information and translate it into IP packets.
Additionally, the gateway may need to use techniques similar to Network Address Translation
(NAT) to decide which source address to use when sending packets to the Internet.

 Internet to WSN Communication

The scenario where a remote Internet-based entity tries to access services from a WSN is more
complex. If the remote terminal has direct communication capabilities with the WSN (for
example, via a compatible transceiver), no special treatment is required. However, if the remote
terminal cannot directly communicate with the WSN, the assistance of a gateway node is needed.

The first step is service discovery: how does the remote terminal know about the existence of the
WSN and its gateway? Once this information is obtained, accessing services in the WSN
becomes a challenge because the WSN typically uses data-centric communication models, not
individual node addressing. In other words, it’s difficult to address a specific sensor node
because the focus is on the data rather than the nodes themselves.

In this case, the remote terminal can send a request to the gateway, which acts as an application-
level gateway or proxy. The gateway translates the request into the necessary protocol for
interacting with the WSN. It masks the differences between the data-centric exchange of the
WSN and the identity-centric exchange of the Internet. The remote terminal does not need to
maintain multihop routes or handle network-specific issues; the gateway simplifies these
complexities.

One possibility is to use Web Service Protocols like WSDL (Web Service Description
Language), which can describe the WSN's services and make it easier for remote users to
interact with the WSN via standard web services. However, this area is still under active
research, and there is no universal standard yet.

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 9

  WSN Tunneling

Gateways can also serve as extensions of one WSN to another, creating a larger "virtual" WSN.
This is achieved by tunneling protocol messages between two networks and using the Internet as
a transport network. This approach allows separate WSNs to be linked together seamlessly,
though it is important to distinguish between virtual links created by gateways and real physical
links. Misunderstanding this distinction can lead to confusion in protocols that rely on physical
properties, such as time synchronization or localization protocols.

In some cases, even mobile nodes can act as intermediaries to connect different WSNs,
providing an additional layer of flexibility. For instance, a mobile user carrying a device could
help bridge two sensor networks that are not directly connected.

This concept of tunneling provides a flexible solution for scaling and interconnecting sensor
networks, although care must be taken to ensure that the physical properties of each WSN are not
disrupted by the virtual connections.

Dr. Vijay Kumar Gowda B N, Assistant Professor, ECE, BGSIT Page 10