Computational intelligence based data aggregation technique in clustered wsn

Computational Intelligence Based Data
Aggregation Technique in Clustered WSN:
Prospects and Considerations
Muhammad Umar Farooq

Proceedings of the World Congress on Engineering and Computer Science 2012 Vol II
WCECS 2012, October 24-26, 2012, San Francisco, USA

Keywords—Data Aggregation, Computational Intelligence,
Clustering, Wireless Sensor Networks, Routing

Reporter

NILAMADHAB MISHRA
D0121008

Contents
• Research Intention
• Preparatory Concept
• WSN network architecture and routing
protocols
• Data aggregation and fusion in WSN
• Data aggregation protocols based on network
architecture
• Computational intelligence and data
aggregation in sensor networks
• Conclusion
2

Research Intention
• Data aggregation in wireless sensor networks is a hot research
focus in recent times.
• Data aggregation is defined as the process of aggregating the
data from multiple sensors to eliminate redundant
transmission and provide fused information to the base
station.
• The main goal of data-aggregation algorithms is to gather
and aggregate data in an energy efficient manner so that
network lifetime is enhanced.
• The data aggregation helps to improve the performance of
WSN routing protocol and further improves the overall
performance of the sensor networks.
• Hierarchical network or clustering plays vital role for data
aggregation ,where sensor nodes are divided into groups and
assigned various roles to each group.

3

Research Intention
• So better clustering of sensor nodes => better data
aggregation/fusions => better performance in WSN routing =>
better life time of sensor networks or better network
longevity.
• Hence in order to improve the efficiency of data aggregation
technique , the author’s intentions to integrate the
computational intelligence techniques with the conventional
data aggregation techniques for making the solution non-
conventional/non-traditional.
• The computational intelligence techniques comprise artificial
neural computing ,fuzzy computing ,evolutionary computing,
swarm intelligence and their novel integrations.
• Hence should use a novel hybridization of computational
intelligence techniques for effective and efficient data
aggregation.

4

Contents
protocols
architecture
• Conclusion
5

Preparatory Concept
• A sensor node consists of mainly 3 subsystems as follows.
*sensor subsystem performs sensing function in the area where it is deployed.
*processing subsystem performs obligatory computations at sensor node level.
*communication subsystem disseminates the sensed data across the sensor networks.
• The data dissemination from source to sink Usually requires the transit node
positioning Information either through GPS-free localization or relative localization or
absolute Localization techniques.

6

Preparatory Concept

The data aggregation usually involves the fusion of data from multiple sensors at intermediate
nodes and transmission of aggregate data to the sink or base station. The data aggregation
attempts to collect the most critical data from sensors and make it available to the sink in low
latency(potential) and energy efficient manner.
So the data aggregation techniques can be used to combine several correlated data signals
into a smaller set of information that maintains the effective data of the original signals.
7

Contents
protocols
architecture
• Conclusion
8

WSN network architecture and routing protocols

• The network structures in WSN can be classified as flat
networks, hierarchical networks and location based networks
and deal with routing differently.
• For routing in a flat network, the nodes are same, have same
capabilities and similar tasks.
• In case of hierarchical networks, the sensor nodes are divided
in groups called clusters.
• In case of location based networks, the locations of the
sensor nodes are used in calculating and selecting a route
from source to the destination.
• In a data-centric network, long-term storage (e.g. disk
storage) of the data is also a part of the "network" and may
serve requestors of that data in the future.

9

• The mobility based networks exploit mobility for data
collection in WSNs. It considers properties of sink mobility as
well as the wireless communication methods for data
transfer:
Mobile base station (MBS)-based solutions:
• An MBS is a mobile sink that changes its position during
operation time. Data generated by sensors are relayed to MBS
without long term buffering.
Mobile data collector (MDC)-based solutions:
• An MDC is a mobile sink that visits sensors. Data are buffered
at source sensors until the MDC visits the sensors and
downloads the information over a single-hop wireless
transmission.

10

• Multipath based networks

11

Heterogeneity based networks
• A heterogeneous WSN consists of three types of nodes: sink node, high-
end senor node (NH), and low-end senor node (NL). Each node has the
same communication model and two types of sensor nodes have the same
sensing model. The difference between NH and NL is that the different
pre-defined communication and sensing range.
• Heterogeneous wireless sensor network (heterogeneous WSN) consists of
sensor nodes with different ability, such as different computing power ,
sensing range and communication range.
In wireless sensor networks, cooperative routing protocols have found good
utility. The nodes in a group or cluster send data to cluster head or sink node,
which aggregates data from many nodes and eliminates the redundant data
at this stage, which ultimately enhances the node and network lifetime.
• Cluster based routing protocols work in two levels mostly, cluster heads are
selected in first level and the second level is used for routing.

composition of one round of the clustering process 12

• Here the author summarized categories of routing protocols for
various sensor networks.

13

Contents
protocols
architecture
• Conclusion
14

Data aggregation and fusion in WSN
• In Data aggregation/collection process data from sensors in an area is
gathered (for performing fusion) to eliminate redundant data and provide
only the useful fused /integrated information to the base station.
• The sensor nodes send their data to the intermediate nodes for data
aggregation which also normally includes the fusion process and then
transmit the condensed data the base station or sink node.
• The aim of data aggregation is to get the most relevant data from the data
collected by the sensor nodes and make it available to the base station
with minimum energy consumption and also with minimum delay.
• Data aggregation mechanism is an important factor in the energy efficient
sensor network along with network architecture and the underlying
routing protocol.

Data aggregation 15

• The Data aggregation techniques include data centric and address centric
routing heuristics. The making of an optimal aggregation tree is NP-hard,
but there exist some sub optimal solutions :
• Center at Nearest Source Data Centric (CNSDC): In CNSDC, the node
closest to the sink acts as the aggregator and the remaining nodes send
their data to that node.
• Shortest Path Tree Data Centric (SPTDC): A SPC is formed and merges the
shortest paths from each source wherever these overlap in an
opportunistic manner.
• Greedy Incremental Tree Data Centric (GITDC): The algorithm starts with
the path from sink to the nearest source, and sequentially adds the next
nearest source to the existing tree.
• Address Centric (AC): In address centric no data aggregation is performed
and all the nodes communicate directly with the sink by using the shortest
paths available.
• Data aggregation has proven to save energy in many network applications.
But data aggregation results in increased delay due to in network
processing, so efficient routing algorithms may be designed to
accommodate for this.

Data aggregation
16

• In data fusion at a sensor node ,the data aggregated from multiple nodes is
processed in such a way that the data generated after fusion has reduced the
transmission overhead in the network.
• Various Conventional techniques are used for data
fusion include Kalman filter, Bayesian networks and
Dempster-Shafer .
• The data fusion process includes the collection of data at the fusion node from
other nodes and fusing the sensed data with its own on the basis of pre
decided criterion.
• Data fusion criteria can be voting, probability-based Bayesian model and stack
generalization and the voting scheme can be based on majority voting,
complete agreement and weighted voting.
• After fusion it transmits the data to another node or directly to the base
station. The important considerations in data fusion include the reporting
time, fusion criteria and the data fusion architecture.

Data fusion 17

data fusion architecture:-
• In centralized architecture, the fusion process is performed at a central node.
– Advantage: Erroneous report(s) can be easily detected.
– Disadvantage: inflexible to sensor changes and the workload is concentrated at a
single point.

• In decentralized architecture, each node performs data fusion itself and no
gathering of data at a fusion node is required.
– Data fusion occurs locally at each node on the basis of local observations and the
information obtained from neighboring nodes.
– No central processor node.
– Advantage: scalable and tolerant to the addition or loss of sensing nodes or
dynamic changes in the network.

• In hierarchical architecture, the nodes are divided into clusters or levels, and
nodes send their data to the higher level nodes for fusion.
– Nodes are partitioned into hierarchical levels.
– The sensing nodes are at level 0 and the BS at the highest level.
– Reports move from the lower levels to higher ones.
– Advantage: Workload is balanced among nodes

data fusion architecture 18

Contents
protocols
architecture
• Conclusion
19

Data aggregation protocols based on network architecture
• The architecture of the network influences the performance of the data
aggregation technique being used. The network structures as already
discussed can mainly be divided into Flat networks and hierarchical or cluster
based networks.

20

• The data aggregation techniques designed for flat networks include Push
Diffusion, one phase Pull Diffusion and two phase Pull Diffusion. SPIN (Sensor
Protocols for Information via Negotiation)protocol is the example of push
diffusion, whereas directed diffusion is representative of two phase pull diffusion.
These algorithms improved the performance of the network as compared to
simple flooding in flat networks.
• Hierarchical networks provide the advantages like scalability, energy efficiency
over the flat networks. The hierarchical networks can be divided into different
categories and data aggregation techniques have been proposed for each
category.
The cluster based hierarchical networking and data aggregation protocols for this
category need some unconventional solution from the fields like CI.
LEACH :-LEACH stands for Low-Energy Adaptive Clustering Hierarchy
– The Set-Up Phase
• Where cluster-heads are chosen
– The Steady-State
• The cluster-head is maintained
• When data is transmitted between nodes

21


HEED :- Hybrid Energy Efficient Distributed Clustering
• HEED was designed to select different cluster heads in a field according to the amount of
energy that is distributed in relation to a neighboring node.
• Four primary goals:
– prolonging network life-time by distributing energy consumption
– terminating the clustering process within a constant number of iterations/steps
– minimizing control overhead
– producing well-distributed cluster heads and compact clusters.

CLUDDA :- Clustered Diffusion with Dynamic Data Aggregation
• Generally two main phases involved in this kind of protocols are setup phase and steady state
phase.
• The setup phase involves the organization of the network into clusters and the selection of
cluster heads.
• The steady state phase involves data aggregation at the cluster heads and data transmission
between nodes.

22

Contents
protocols
architecture
• Conclusion
23

Computational intelligence and data aggregation in sensor networks

• CI is defined as the computational models and tools of intelligence, that
capable of inputting raw numerical sensory data directly, processing them
by exploiting the representational parallelism and pipelining the problem,
generating reliable and timely responses and withstanding high fault
tolerance.
• Here author proposed Data aggregation and fusion techniques based on
computational intelligence for effective fusion of information gathered
from different sensor nodes, the techniques that can adjust automatically.
• many CI paradigms favor the problems of data aggregation and sensor
fusion , those are evolutionary algorithms GA, fuzzy logic and NNs.
• Efficient data aggregation routes have been achieved by GA in a mobile
agent based WSN because of their inherent parallel nature and ability to
deal with difficult real world problems like non-homogeneous, noisy,
incomplete and/or obscured information.
• Particle swarm optimization PSO can be a good alternate to GA because of
PSO' capability to converge quickly to the optimal solutions and to
determine the optimal route paths.

24

Computational intelligence and data aggregation in sensor networks
• For fusion process neural networks are well suited because neural
networks can learn and dynamically adapt to the changing scenarios.

• Reinforcement learning is fully distributed and it can adapt quickly to
network topology change or any node failure. It has been used efficiently
for finding the optimal path for data aggregation.

• Fuzzy Logic based distributed approach using fuzzy numbers and weighted
average operators to perform energy efficient flooding-based aggregation.

• In wireless sensor networks many situations demand aggregating data at a
central node e.g. monitoring events. For these situations, the centralized
approaches like neural networks, GA or PSO can be used efficiently to
know the features of the data.

25

Contents
protocols
architecture
• Conclusion
26

Conclusion
• Due to the several limitations of wireless sensor network, all
CI techniques can not be applied in the field of data
aggregation and data fusion in literature by researchers, as
their characteristics do not suit data aggregation and fusion.
• Data aggregation needs centralized approaches i.e. the data is
gathered at a central node for performing processing and
transmitting results, it results in increased communication
overhead.
• So distributed data aggregation technique under dynamic
communication constraints may be suggested.
• Future research is likely to focus on developing a well founded
analytical approach to distributed multi-sensor data
aggregation and fusion problems where there are time
varying communication bandwidth constraints.

27

Computational intelligence based data aggregation technique in clustered wsn

More Related Content

What's hot

Similar to Computational intelligence based data aggregation technique in clustered wsn

More from TAIWAN

Computational intelligence based data aggregation technique in clustered wsn