This document summarizes several techniques for achieving efficient data acquisition in wireless sensor networks (WSNs), including an adaptive rate control algorithm, the Enhanced Congestion Detection and Avoidance (ECODA) technique, and the Event-to-Sink Reliable Transport (ESRT) protocol. ECODA uses dual buffer thresholds and weighted buffer differences for congestion detection. ESRT runs primarily on the sink node and aims to configure sensor reporting frequencies to achieve reliable event detection while minimizing energy usage through congestion control. The document evaluates these techniques through simulation and finds they can effectively tackle issues like congestion detection/avoidance and controlling fluctuating data reporting rates.
Achieving Efficient Data Acquisition in Wireless Sensor Networks
1. 81 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
Achieving Efficient Data Acquisition Techniques in Wireless
Sensor Networks
Rutvik K. Pensionwar1
, Nilesh Thite2,
Pranav Tambat3
,
Onkar Tummanpalli4
,Sambhaji S. Sarode5
Computer Engineering Department,
MIT-College of Engineering, Pune-38,
Maharashtra, India.
ABSTRACT:
For a WSN, acquisition of data, storage in the
secondary memory and working on the
protocols across various layers is the principle.
But maximizing the lifetime, uniform data
collection, and taking care of reliability and
congestion detection-control becomes critical.
In a WSN it may happen that a sensor node
fails to report the readings to the sink each and
every time it senses/stores the data. The
solution of the problem lies beneath the
development of efficient data reporting strategy
with high accuracy. So, this paper deals with an
adaptive rate control algorithm so as to achieve
near to ideal data reporting rate under the
bandwidth constraint and adaptively adjust the
reporting rate based on residual energy of the
sensor nodes. Another reference is made to
Enhanced Congestion Detection and Avoidance
(ECODA) technique which leads us to energy
efficient and better QoS model by achieving
efficient congestion control and flexible
weighted fairness for different class of traffic.
Also, for an event based system, reliable event
detection at the sink is totally based on
collective information gathered from all the
source nodes. Hence a novel transport protocol
named ESRT is mentioned for reliable event
detection with minimum energy consumption
which mainly runs on sink. Simulation results
show that these major problems like congestion
detection and avoidance, achieving reliability
as well as control over fluctuating data
reporting rate can be tackled using this
protocols to achieve efficient data acquisition in
WSN.
Keywords:
WSN - Wireless sensor Network
ECODA - Enhanced Congestion Detection and
Avoidance
ESRT - Event to Sink Reliable Transport
QoS - Quality of Service
H-B-H - Hop by Hop
E-T-T - End to End
NS-2 - Network Simulator v2
1. INTRODUCTION:
From last two decades there have been an
increase in research and development of
Wireless Sensor Networks, a key technology
for various applications that involves low cost
and long term monitoring such as variety of
environmental monitoring applications like
habitat monitoring, snow monitoring,
temperature monitoring, civil structures
monitoring and noise pollution monitoring,
flood monitoring. A common aim of a WSN is
to sense/store the data and report that collected
data to the sink. Hence, battery as a power
source plays a vital energy media of the sensor
node which is expected to work for a year or
two or more and that to w/o replenishing.
Therefore, energy efficiency becomes critical
issue. Along with that,there are certain issues
that we should be concern with like congestion
detection, control and avoidance as well as
reliable event detection and control over data
reporting rate. When the events are detected,
the information stored in the secondary
2. 82 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
memory is of great importance. However, the
busy traffic resulted due to detected events can
easily cause congestion. Along with congestion,
reliability is another factor which should be
taken care of. Studies show thatreporting rate of
data affects the reliability of the system and
hence should be controlled efficiently. The
results of these issues must be foreseen and
prevention against them must be taken. Hence,
we will refer to the profound protocols to avoid
such situations.
In WSN, there are two types of
congestion viz., TransientCongestion, which is
caused due to link variation and second one is
persistent congestion caused by variation in
source data sending rate. For transient
congestion, each sensor monitors channel
utilization and buffer occupancy level to detect
congestion whereas for persistent congestion,
source requires sink’s feedback to maintain its
data rate.
Similarly, the congestion control is
roughly classified into two mechanisms viz.,
end-to-end congestion control and hop-by-hop
congestion control. End-to-End congestion
control performs exact rate adjustment at
source and coordinatornodes but runs with a
drawback that it relies completely in round-trip
time which makes the process slow and time
consuming. On the contrary, the Hop-by-Hop
congestion control has faster response.
Rather than discussing various types of
congestion we should focus on their solutions.
Hence, we need to consider the fairness issue
for packets with different importance or
priority. So, we refer to an efficient congestion
detection and avoidance (ECODA) protocol
[1], which has a distributed mechanism
operating at the network and MAC layer. Here,
following ideas are incorporated:
1.ECODA uses dual buffer thresholds
and weighted buffer difference for congestion
detection. The method differs from traditional
single buffer threshold method.
2. It adopts a novel method to filter packets
according to channel loading and packets
priority when congestion happens.
3. For transient congestion, H-B-H implicit
backpressure manner is used whereas for
persistent congestion, bottleneck node based
source sending rate control and multi-path
loading balancing are proposed.
Congestion detection and avoidance
(CODA) protocol [2], a predecessor of ECODA
differs from it a bit. CODA comprises of three
mechanisms viz.,
1. Receiver based congestion detection and
control,
2. Open loop hop-by-hop backpressure and
3. Closed loop multi-source regulation.
Along with this, a reference is also made to a
novel transport solution named Event-to-Sink
Reliable Transport (ESRT) [3], developed to
achieve reliable event detection together with
congestion resolution functionality. Some of
the silent features of ESRT are as following:
1. Self-Configuration: Under dynamic
topologies, ESRT configures itself and achieves
flexibility by self-adjusting the operating point
(discussed further).
2. Energy Awareness: Energy plays a vital
role in the lifetime of a network and hence
ESRT takes care of it by adjusting the data
reporting rate dynamically.
3. Congestion Control: ESRT owns an
important feature that congestion control is also
used to reduce energy consumption. A suitable
congestion control mechanism can help
conserve energy while maintaining desired
accuracy levels at sink by conservatively
reducing the reporting rate.
4. Collective Identification: ESRT does not
require individual node ID’s for the operation
since the ESRT runs only on sink and sink is
only interested in collective information
provided by numerous sensor nodes rather
individual reports.
5. Biased Implementation: Running ESRT
only on sink helps conserve energy and thus
3. 83 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
own a power to shift the burden to the high
powered sink. This can only be made possible
with the help of ESRT.
The rest of the paper is organized as follows:
After listing protocols for each major problem,
we will focus on their definition and algorithm
implementation individually. Each problem will
be briefly serviced and simulation results will
be drawn respectively.
2. ECODA PROTOCOL OVERVIEW:
The ideas incorporated in ECODA are briefed
as follows:
1. Use of dual buffer thresholds and
weighted buffer difference for congestion
detection
Here, dual buffer threshold is adopted to
measure congestion and hence it consist of two
threshold values (Qmax and Qmin) and hence
three states of buffer are formed viz., accept
state, filter state and reject state. Even though
this method along with data priority is
important in congestion control, they are not
adopted much. Hence, according to [1], the
obvious choice is practicing weighted buffer
difference in the neighborhood as congestion
resolving sequence basis.
Fig Buffer State
2. Flexible Queue Scheduler and weighted
fairness
For a tree structural sensor network, there’s
unfairness in bandwidth allocation due to
variation in priorities as like more priority
given to nodes to send their packets near sink
and less to rest of them. Hence, a Flexible
Queue Scheduler is proposedwhere two priority
factors are defined: static and dynamic, where
static priority is an integer value and dynamic
priority is drawn as below:
DP (packet) = [α*hop + SP (packet)] / [1+
β*delay]
Where:
α and β - parameters for tuning performance,
hop - number of hops to sink,
SP - Static priority of packet
delay - time from packet generation to
current
Hence, packet dynamic priority is partially
determined by the number of hops to sink and
delay. With increasing hop number, DP
increases and coverage fidelity is guaranteed.
So, Qmin and Qmax are used for handling packets
with different strategies. (For N= no. of
packets)
1. If 0≤N≤Qmin, all the incoming packets are
buffered, since queue utilization is low.
2. If Qmin≤N≤Qmax, some low dynamic priority
packets are dropped or overwritten by high
dynamic priority, and average buffer length
increases at rate of about R≤∂1.
3. If Qmax≤N≤Q, some high dynamic priority
packets are dropped or overwritten, average
buffer length increases at rate of about R≤∂2.
∂1 and ∂2 are two variables tuned to achieve
optimal system performance. The below
simulation results shows that:
a) When buffer length is between Qmin and
Qmax, bigger the value of ∂1, buffer length
increases at faster rate and packet dropping rate
lowers.
b) When buffer length is between Qmax and Q,
buffer utilization is high and the value of ∂2
should be lesser than ∂1 so as to reject more
packets to lower down buffer increasing rate.
c) When the packets have higher dynamic
priority, ∂1 and∂2 are set relatively large to
accept important packets, else set relatively low
to discard less important packets.
4. 84 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
Fig Buffer Changing Rate
Fig Packet Dropping Rate
3. Bottleneck-node based source sending rate
control and multi-path loading balancing
The method is so proposed that every node can
determine the routing path by itself along with
adjusting the source data sending rate more
accurately and efficiently.
3. THE RELIABLE TRANSPORT
PROBLEM IN WSN:
Consider a WSN where sink decides on the
event features every ‘T’ time units. That means
‘T’ is the decision interval where sink makes an
informed decision based on the reports received
from the sensor nodes during that interval.
Also, we assume that sink derives a reliability
indicator ri at the end of decision interval.
1. The observed reliability, ri is the number of
received data packets in decision interval ‘i’ at
the sink.
2. The desired event reliability, R is the number
of data packets required for reliable event
detection.
Based on the comparison of values of ri and R,
appropriate actions are taken to achieve the
reliability. The reporting rate of the sensor node
is defined as the number of packets sent out per
unit time by that node. The transport problem in
WSN is to configure the data reporting rate f of
the sensor nodes so s to achieve the required
event detection reliability, R at the sink with
minimum resource utilization.
In order to study the relationship between these
parameters an artificial environment was
created and results were plotted using ns-2.
Fig Effect of varying the reporting rate, f, of
source nodes on the event reliability, ri,
observed at the sink. The number of source
nodes is denoted by n.
5. 85 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
The following was observed:
1. The reliability, r, shows a gradual increase
with source reporting rate, f, until a certain f =
fmax, beyond which the reliability drops. This is
because the network is unable to handle the
increased load of data packets and packets are
dropped due to congestion.
2. Such an initial increase and subsequent
decrease in reliability is observed regardless of
the number of source nodes, n.
3. fmax decreases with increasing n, i.e.,
congestion occur at lower reporting frequencies
with greater number of sources.
4. For f > fmax, the behavior is not smooth. An
intuitive explanation for such a behavior is as
follows. The number of received packets,
which is our reliability, r, is the difference
between the total number of source data
packets, s, and the number of packets dropped
by the network, d. While s simply scales
linearly with f, the relationship between d and f
is non-linear. In some cases, the difference s-d
is seen to increase even though the network is
congested. The important point to note
however, is that this wavy behavior always
stays well below the maximum reliability at f =
fmax
5.The drop in reliability due to network
congestion is more significant with increasing
n.
The five characteristic regions: (η: normalized
measure of reliability = r/R)
1. (NC, LR): f < fmaxand η< 1-ϵ (No
Congestion, Low Reliability)
2. (NC, HR): f ≤ fmax and η > 1+ϵ (No
Congestion, High Reliability)
3. (C, HR): f > fmax and η > 1 (Congestion,
High Reliability)
4. (C, LR): f > fmax and η ≤ 1 (Congestion,
Low Reliability)
5. OOR: f < fmaxand 1-ϵ ≤ η ≤ 1+ϵ (Optimal
Operating Region)
Fig the five characteristic regions in the
normalized event reliability η versus
reporting frequency, f, behavior
4. ESRT: EVENT-TO-SINK RELABLE
TRANSPORT PROTOCOL
The primary motive of ESRT [3] is to achieve
and maintain operation in state OOR. Hence,
the aim is to configure the reporting frequency f
to achieve the desired event detection accuracy
with minimum energy expenditure. To help
accomplish this, ESRT uses a congestion
control mechanism that serves the dual purpose
of reliable detection and energy conservation.
In general, the network can reside in any one of
the 5 states Si ϵ {(NC, LR), (NC, HR) (C, HR)
(C, LR), OOR}. Depending on the current state
Si, ESRT calculates an updated reporting
frequency fi+1 which is then broadcasted to
source nodes.
The algorithms of ESRT mainly run on the
sink, with minimal functionality at the source
nodes. More precisely, sensor nodes only need
the following two additional functionalities
viz.,
1. Sensor nodes must listen to the sink
broadcast at the end of each decision interval
and update their reporting rates
6. 86 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
2. Sensor nodes must deploy a simple and
overhead-free local congestion detection
support mechanism
Such a graceful transfer of complexity from
sensor nodes to the sink node reduces
management costs and saves on valuable sensor
resources.
ESRT identifies the current state Si from
1. Reliability indicator ηi computed by the sink
for decision interval i
2. A congestion detection mechanism,
Using the decision boundaries previously
defined. Depending on the current state Si, and
the values of fi and ηi, ESRT then calculates the
updated reporting frequency fi+1 to be broadcast
to the source nodes. At the end of the next
decision interval, the sink derives a new
reliability indicator ηi+1 corresponding to the
updated reporting frequency fi+1 of source
nodes. In conjunction with any congestion
reports, ESRT then determines the new network
state Si+1. This process is repeated until the
optimal operating region is reached.
ESRT also adopts a frequency update policy for
each and every state based on which the
algorithm of ESRT protocol operation is put
forward.
k = 1;
ESRT()
If (CONGESTION)
If (η < 1)
/* State=(C,LR) */
/* Decrease Reporting Frequency
Aggressively */
f = f η/k
;
k = k + 1;
else if (η > 1)
/* State=(C,HR) */
/* Decrease Reporting Frequency to
Relieve Congestion; No Compromise on
Reliability */
k = 1;
f = f / η ;
end;
else if (NO CONGESTION)
k = 1;
If (η < 1 - ϵ )
/* State=(NC,LR) */
/* Increase Reporting Frequency
Aggressively */
f = f / η ;
else if (η > 1 + ϵ )
/* State=(NC,HR) */
/* Decrease Reporting Frequency
cautiously */
f = f/2 (1 + (1/ η));
end;
else if (1-ϵ ≤ η ≤ 1+ ϵ)
/* Optimal Operating Region */
/* Hold Reporting Frequency */
f = f;
end;
end;
Algorithm of the ESRT protocol operation.
5 MEASURES FOR RATE-CONSTRAINT
UNIFORM DATA COLLECTION IN WSN
As we have discussed earlier that it is not every
time a WSN node would report the readings to
the data processing center due to limited
network resources or some other reasons. So,
two different data reporting strategies are
composed viz. prediction-based reporting and
temporal-based reporting.
Prediction Based Reporting:
This effective technique for data reporting has a
basic ideology that if the sensor itself can
predict the reading at the sink node, then it is
not mandatory to report the data at the server.
An online scheme is proposed based on
compression and prediction for construction of
approximate time series which would guarantee
less energy consumption. An adaptive as well a
lightweight prediction model selection scheme
is proposed for sensor node to automatically
determine a statistically good performance
model amongst many.
7. 87 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
Temporal Based Reporting:
It is another efficient data reporting strategy
after prediction-based reporting where a sensor
node does not report to the sink if it does not
deviate from the last reported readings beyond
an error bound as well going with an
assumption that all unreported readings remain
unchanged. Two different strategies are
composed on behalf of this reporting strategies
viz. one aims to improve network performance
i.e. the lifetime of a network with lowest
possible requirement on data quality whereas
other is a vice-versa i.e. improving data quality
subject to bandwidth.
Our objective is to find the rationale behind this
equi-interval strategy; we have to consider one-
hop transmission for data reporting so that the
server node is the neighbor of each and every
sensor node.
Here, we introduce the concept of Adaptive rate
control algorithm from previous studies [4].
Data reporting is the vital part in WSN of
which care should be taken. High reporting rate
leads to network congestion whereas low rate
results in low bandwidth utilization. So, these
both cases should be avoided to play safe
Algorithm 2:
Centralized Rate Adaption
Input:
{Pdrop}: packet drop rate vector
{Qbuf}: buffer occupancy vector
Output:
ƛ: updated reporting rate
1. for each update each period Tupd do
2. calculate avg({Pdrop}) and ({Qbuf});
3. if avg({Pdrop}) ≥ Pmax then
4. ƛ = ƛ / (1 + avg({Qbuf}) / Lbuf)2
;
5. broadcast updated ƛ;
6. else if avg({Pdrop}) < Pmin holds for 3Tupd
then
7. ƛ = ƛ + ∂;
8. broadcast updated ƛ;
9. end if
10. end for
In the above algorithm, each sensor node
calculates its average buffer occupancy Qbuf
over a certain period and encapsulates this
value into each outgoing packet. Qbuf is
calculated as follows: Qbuf = (1 - ω)Qbuf + ω q,
where ω is a weight coefficient and q is the
instantaneous buffer occupancy. The sink only
catches the value of the latest received Qbuf for
each sensor. It can calculate the packet drop
rate Pdrop for each sensor node based on the
number of packets received from that node and
the number of packets expected to be received
from that node during the last period Tupd,
based on the reporting rate that it allocated last
time and the value of Tupd.
At the end of each update period Tupd, the sink
performs this algorithm to adjust the reporting
rate ƛ of each sensor node. In above algorithm,
Lbuf is the buffer size and ∂ is the step-size of
rate increment.
In algorithm, when avg({Pdrop}) ≥ Pmax, it
indicates that ƛ is higher than a given upper
threshold on the transmission rate in the
network and the sensors need to decrease their
reporting rate. In some AIMD strategies, the
multiplicativedecreasing factor is constant, that
is, using ƛ = ƛ /h (h is a constant) for rate
decreasing. On one hand, if h is large, ƛ,which
is slightly higher than the upper threshold, may
become too small after rate decreasing. On the
other hand, if h is small, it may need a long
time to decrease ƛ that is far beyond the upper
threshold. For adaptive rate control, in above
algorithm, ƛ is decreased more quickly if Qbuf is
higher, and is decreased relatively more slowly
if Qbuf is lower.
When avg({Pdrop}) < Pmin, the bandwidth
utilization is mostly insufficient and the sensors
should increase their reporting rates. However,
to avoid blind rate increasing, we need to
estimate the performance of the network in a
steady state. That is why we require that
avg({Pdrop}) < Pmin hold for 3Tupd.
8. 88 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
Once ƛ is updated, each node adjusts its
reporting rate to ƛ; the sink resets {Pdrop} and
each node resets its own Qbuf for parameter
recalculation at the end of next Tupd.
6. EMULATION RESULTS:
Before managing buffer, the packets used to
drop as the number would approach the size of
buffer. The reason was the insufficient time for
the buffer manager to clear up the queue and
hence the packet arriving after the count of size
of buffer used to drop. We performed a
practical demonstration on JN-5139 devices
and managed the buffer in a way that
minimized the chances of packet drop. The
average packet drop was minimized by about
85 %. When there should have been a complete
drop of about packets, the model seen about 85-
90 % increase in packet delivery.
Fig. The size of group of packets at each
interval is 100 and fig shows no. of packets
received for continuous time interval.
Also, we have plotted the simulation results
(demonstration on JN-5139) regarding data
frame transfer time and acknowledgement time
(in milliseconds) with respect to packet number
(only 5 readings included).
7. CONCLUSION
In a WSN, along with an energy efficient
communication we also require a congestion
free network as well as a data reporting strategy
that will help us achieve data accuracy. In this
paper we presented a congestion control
protocol (ECODA) that detects congestion
using dual buffer thresholds and weighted
buffer difference, which further deals with
transient and persistent congestion to achieve
high link utilization, flexible fairness, reduction
in packet loss and lowering of delay. We also
studied ESRT, a novel transport solution whose
congestion control component serves the dual
purpose of achieving reliability and conserving
energy. The primary objective of ESRT is to
configure the network as close as possible to
optimal operating point (OOP), where
reliability is achieved with minimum energy
consumption and w/o network congestion.
Finally, we proposed an adaptive rate control
algorithm to achieve maximum data reporting
rate under bandwidth constraint and further
extend this to one that can adaptively adjust the
reporting rate of a sensor, based on residual
9. 89 Rutvik K. Pensionwar, Nilesh Thite, Pranav Tambat, Onkar Tummanpalli,Sambhaji S. Sarode
International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 4, Issue4
April 2015
energy of the sensor to prolong the network
lifetime.
7 REFERENCES
[1]ECODA: Enhanced Congestion Detection
and Avoidance for Multiple Class of Traffic
in Sensor Networks by Li Qiang Tao, Feng
Qi Yu
[2]CODA:Congestion Detection & Avoidance
in Sensor Networks byChieh-Yih Wan,
Shane B. Eisenman, Andrew T. Camp
[3]Event-to-Sink Reliable Transport in
Wireless Sensor Networks byÖzgür B.
Akan, Ian F. Akyildiz
[4]Rate-constrained uniform data collection in
wireless sensor networks byH. Deng, B.
Zhang, J. Zheng
[5]Priority Enabled Transport Layer Protocol
for Wireless Sensor Network byAtif Sharif,
Vidyasagar Potdar, A.J.D.Rathnayaka.