V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.

2
ISSN: 0975-5665, May, 2013 www.ijrct.org

Tree-Based Wireless Sensor Networks in Fast Data Collection

#V.REDYA JADAV1, Associate Professor, Computer Science and Engineering Department

#D.SRAVAN2, Assistant Professor Computer Science and Engineering Department
#A.DEVILAL3M.Tech, Computer Science and Engineering Department
# Bomma Institute of Technology and Science, Khammam, A.P State, INDIA

of the routing tree. To this end, we construct

degree-constrained spanning trees and
Abstract: capacitated minimal spanning trees, and
We investigate the following fundamental show signicant improvement in scheduling
question - how fast can information be performance over different deployment
collected from a wireless sensor network densities. Lastly, we evaluate the impact of
organized as tree? To address this, we different interference and channel models on
explore and evaluate a number of different the schedule length.
techniques using realistic simulation models In this paper, a new distributed power-
under the many-to-one communication control (DPC) scheme is suggested to
paradigm known as converge cast. We rst improve convergence speedand system
consider time scheduling on a single robustness against carrier-to-interference-
frequency channel with the aim of ratio (CIR) estimation errors. To expedite
minimizing the number of time slots the CIR balancing in our DPC scheme, an
required (schedule length) to complete a instability detection rule was used. As
converge cast. Next, we combine scheduling compared with Foschinis DPC (FDPC)
with transmission power control to mitigate method, numerical results indicated that the
the effects of interference, and show that proposed algorithm achieves performance
while power control helps in reducing the improvements in terms of outage probability
schedule length under a single frequency, as well as in the algorithms convergence
scheduling transmissions using multiple speed. More specically, by appropriate
frequencies is more efficient. We give lower selection of some parameters, the algorithm
bounds on the schedule length when speed reduces from about 90 iterations in
interference is completely eliminated, and FDPC to 9 iterations in the proposed
propose algorithms that achieve these algorithm. The system robustness against
bounds. We also evaluate the performance CIR estimation errors was also explored.
of various channel assignment methods and Index Terms:-Cellular systems, CIR
nd empirically that for moderate size balancing, distributed algorithms, instability
networks of about 100 nodes, the use of detection, power control, Converge cast,
multi-frequency scheduling can sufce to TDMA scheduling, multiple channels,
eliminate most of the interference. Then, the power-control, routing trees.
data collection rate no longer remains
limited by interference but by the topology

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

1 INTRODUCTION:

Frequency reuse in wireless cellular and N is the number of source nodes. We

networks results in increased system then introduce optimal time slot assignment
capacity. It increases, however, the co schemes under this scenario which achieve
channel interference that imposes limitations these lower bounds.
on the minimum reuse distance. An ecient Evaluation of Power Control under
technique in reducing the co channel Realistic Setting: It was shown recently [5]
interference is to control the transmitted that under the idealized setting of unlimited
power in order to provide each receiver with power and continuous range, transmission
a satisfactory reception. A commonly used power control can provide an unbounded
measure of the quality of communications is improvement in the asymptotic capacity of
the carrier-to-interference ratio (CIR) at the aggregated converge cast. In this work, we
receiver. evaluate the behavior of an optimal power
Converge cast, namely the collection of data control algorithm [6] under realistic settings
from a set of sensors toward a common sink considering the limited discrete power levels
over a tree-based routing topology, is a available in todays radios. We nd that for
fundamental operation in wireless sensor moderate size networks of 100 nodes power
networks (WSN) [1]. In many applications, control can reduce the schedule length by 15
it is crucial to provide a guarantee on the 20%.
delivery time as well as increase the rate of Evaluation of Channel Assignment
such data collection. For instance, in safety Methods: Using extensive simulations, we
and mission-critical applications where show that scheduling transmissions on
sensor nodes are deployed to detect oil/gas different frequency channels is more
leak or structural damage, the actuators and effective in mitigating interference as
controllers need to receive data from all the compared to transmission power control. We
sensors within a specic deadline [2], failure evaluate the performance of three different
of which might lead to unpredictable and channel assignment methods: (i) Joint
catastrophic events. Frequency and Time Slot Scheduling
The following lists our key ndings and (JFTSS), (ii) Receiver-Based Channel
contributions: Assignment (RBCA) [7], and (iii) Tree-
Bounds on Converge cast Scheduling: Based Channel Assignment (TMCP) [8].
We show that if all interfering links are These methods consider the channel
eliminated, the schedule length for assignment problem at different levels: the
aggregated converge cast is lower bounded link level, node level, or cluster level. We
by the maximum node degree in the routing show that for aggregated converge cast,
tree, and for raw-data converge cast by TMCP performs better than JFTSS and
max(2nk1,N), where n k is the maximum RBCA on minimum-hop routing trees, while
number of nodes on any branch in the tree, performs worse on degree-constrained trees.

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

For raw-data converge cast, RBCA and than 500 samples per second are required to
JFTSS perform better than TMCP, since the efficiently detect damages [2]. Time division
latter suffers from interference inside the multiple access (TDMA) scheduling is a
branches due to concurrent transmissions on natural solution for such periodic data
the same channel. collection applications [3], [4]. Consider a
Impact of Routing Trees: We investigate repeated frame of k time slots in which each
the effect of network topology on the link of the data gathering tree is scheduled
schedule length, and show that for once. In steady state (once a pipeline is
aggregated converge cast the performance established), the sink will receive aggregated
can be improved by up to 10 times on information from all nodes in the network
degree constrained trees using multiple once per frame, i.e. once every k slots. In
frequencies as compared to that on this framework, maximizing the data
minimum-hop trees using a single collection rate corresponds exactly to
frequency. For raw-data converge cast, minimizing the frame length. This is the
multi-channel scheduling on capacitated focus of our work.
minimal spanning trees can reduce the
schedule length by 50%. We explore a number of techniques in order
Impact of Channel Models and to address the basic question: how fast can
Interference: Under the setting of multiple aggregate data be streamed to the sink?
frequencies, one simplifying assumption These techniques provide a hierarchy of
often made is that the frequencies are successive improvements. The simplest
orthogonal to each other. We evaluate this approach is to do some form of interference-
assumption and show that the schedules aware minimum frame-length TDMA-
generated may not always eliminate scheduling that enables spatial reuse. The
interference, thus causing considerable second step is to combine the scheduling
packet losses. We also evaluate and compare with transmission power control. The third
the two most commonly used interference step is to consider the use of multiple
models: (i) the graph-based protocol model, frequency channels. We show that once
and (ii) the SINR (Signal-to-Interference- multiple frequencies are employed along
plus-Noise Ratio) based physical model. with spatial-reuse TDMA, the aggregated
data collection rate often becomes no longer
Periodic collection of aggregated data from interference-limited, but rather topology
sensors to a common sink over a tree limited. Thus, the nal step to enhance the
topology is a fundamental operation in rate of periodic aggregated data collection is
wireless sensor networks (WSN). In many to use an appropriate degree constrained tree
such applications, it is of interest to topology. Our primary conclusion is that
maximize the rate at which the sink can combining these techniques can provide an
receive aggregated data from the network order of magnitude improvement in the rate
[1]. For instance, it has been noted that in compared to the simple approach of TDMA
networked structural health monitoring more

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

scheduling on a single channel with clear whether the channels are frequencies,
minimum-hop routing trees. codes, or any other method to eliminate
In the present paper, we present a new DPC interference. Moreover, they assume a
algorithm that achieves CIR balancing with simple interference model where each node
unit probability. Numerical analysis through has a circular transmission range and
computer simulations indicated that our cumulative interference from concurrent
algorithm, in conjunction with the use of the multiple senders is avoided. Different from
instability detection rule, has excellent their work, we consider multiple frequencies
performance compared to Foschinis DPC and evaluate the performance of three
algorithm (FDPC) in terms of outage different channel assignment methods
probability and convergence speed. together with evaluating the effects of
transmission power control using realistic
2. RELATED WORKS: interference and channel models, i.e.,
physical interference model and overlapping
We experimentally investigated the impact channels and considering the impact of
of transmission power control and multiple routing topologies. Song et al. [12] extended
frequency channels on the schedule length; their work and proposed TDMA based MAC
we proposed constant factors and protocol for high data rate WSNs in [16].
logarithmic approximation algorithms on Tree MAC considers the differences in load
geometric networks (disk graphs). at different levels of a routing tree and
Frequency reuse in wireless cellular assigns time slots according to the depth, i.e.
networks results in increased system the hop count, of the nodes on the routing
capacity. It increases, however, the co tree, such that nodes closer to the sink are
channel interference that imposes limitations assigned more slots than their children in
on the minimum reuse distance. An ecient order to mitigate congestion. However, Tree
technique in reducing the co channel MAC operates on a single channel and
interference is to control the transmitted achieves 1/3 of the maximum throughput
power in order to provide each receiver with similar to the bounds presented by Gandham
a satisfactory reception. A commonly used et al. [1] since the sink can receive every 3
measure of the quality of communications is time slots. The problem of minimizing the
the carrier-to-interference ratio (CIR) at the schedule length for raw-data converge cast
receiver. on single channel is shown to be NP-
For raw-data converge cast, Song et al. [12] complete on general graphs by Choi et al.
presented a time-optimal, energy-efficient, [13]. Maximizing the throughput of
packet scheduling algorithm with periodic converge cast by nding a shortest-length,
trafc from all the nodes to the sink. Once conict free schedule is studied by Lai et al.
interference is eliminated, their algorithm [14], where a greedy graph coloring strategy
achieves the bound that we present here, assigns time slots to the senders and prevent
however, they briey mention a 3-coloring interference. They also discussed the impact
channel assignment scheme, and it is not of routing trees on the schedule length and

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

proposed a routing scheme called disjoint We assume all the nodes in the network are
strips to transmit data over different shortest sources and the data is aggregated such that
paths. However, since the sink remains as the data coming from different sources are
the bottleneck, sending data over different combined into a packet(s) before
paths does not reduce the schedule length. forwarding. If the incoming packets cannot
As we will show in this paper, the be combined in a single packet and multiple
improvement due to the routing structure packets have to be forwarded, we assume
comes from using capacitated minimal each time slot is long enough to transmit
spanning trees for raw-data converge cast, those packets. This is a reasonable
where the number of nodes in a sub tree is assumption since the size of the sensor
no more than half the total number of nodes readings is usually very small. Figure 1
in the remaining sub trees. The use of shows the relationship between the schedule
multiple frequencies has been studied length and the aggregated data rate. The
extensively in both cellular and ad hoc numbers on the links show the assigned time
networks, however, in the domain of WSN, slots and the numbers inside the circles
there exist a few studies that utilize multiple represent the node ids. On the left of the
channels [8], [17], [18]. To this end, we gure we see the schedule showing the
evaluate the efficiency of three particular received packets from the associated senders
schemes that treat the channel assignment at by each parent on each time slot. After
different levels. frame 1, once the sink gets initial data from
each source (a pipeline is established), the
3. MECHANISMS same schedule is repeated and the sink
collects the aggregated data from the
A. Preliminaries network at a rate of 3 time slots. Thus, the
Before explaining the studied mechanisms, schedule length should be minimized to
we rst express the preliminary design improve the data collection rate.
details and assumptions:
We consider a static wireless sensor
network. The sensor nodes periodically
sense the environment and send their
readings over a multi-hop tree topology.
Time is divided into equal sized slots that
are grouped into frames. We focus on
minimizing the length of the frame such that
each node is assigned one time slot.
We consider minimum-hop routing trees
where all the nodes select a parent node B. Joint Scheduling and Power Control:
where they transmit their readings to be
forwarded towards the sink node. A cross layer method for joint scheduling
and power control in wireless multi-hop

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

networks. They proposed an optimal

distributed algorithm to improve the C. Frequency and Time Scheduling
throughput capacity of wireless networks.
The aim is to nd a TDMA schedule which The use of multiple frequency channels is an
can support as many transmissions as efficient way to improve the capacity of
possible in each time slot. We use their wireless networks. Simultaneous
algorithm to investigate the impact of power transmissions on different frequencies (if the
control on the scheduling performance. frequencies are not orthogonal, different
The solution is composed of 2 phases: frequencies may also be conicting. We use
scheduling and power control. It is to be non-conicting frequencies and different
executed at the beginning of each time slot frequencies interchangeably in the text) can
in order to cope with excessive interference take place without interference in the same
levels. The scheduling phase searches for a spatial neighborhood.
transmission schedule which is dened to be In this section we introduce a simple
valid if no node is to transmit and receive scheduling method which separately assigns
simultaneously and no node is to receive the time slots and frequencies on a tree
from more than one neighbor at the same topology. Motivation for this proposal is as
time. Power control phase iteratively follows:
searches for an admissible schedule which Intersecting links, which are dened as the
means that a set of transmission powers is links with a common destination (Figure 2),
available to satisfy the SINR (signal to cannot transmit on the same time slot since
interference and noise ratio) constraints for they have to wait for each others
all links in the given valid schedule. In each transmission. Assigning non conicting
iteration nodes adjust their transmission frequencies to these nodes does not improve
powers as follows: the situation, either. Then the receiver
= should be assigned a frequency and the
senders should use this frequency to transmit
Where the new transmission power level in to the parent.
the next iteration is, is the current Interfering links are the links which
transmission power level and is the SINR create/face interference if they are scheduled
threshold. If the maximum number of simultaneously. Figure 2 shows an example
iterations is reached and still there are nodes where the dotted line represents interference.
which cannot meet the SINR constraints, Interfering links should not get the same
i.e., if the valid scenario is not admissible, time slot and frequency. Since our aim is to
the scheduling algorithm excludes the link minimize the number of time slots, the best
with the minimum SINR. The power control option then is to assign the same time slot on
algorithm is repeated until an admissible non-conicting frequencies.
transmission scenario is found. Then, the
nodes start transmission using the computed
transmission powers in the current slot.

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

the time slot assignment, the SINR condition

is checked among the interfering senders.
Figure 3 shows a scheduling example on a
tree topology. In the first part of the figure,
The method is composed of 2 phases: the solid lines between the nodes show the
transmission links whereas the dotted lines
frequency scheduling and time slot
show the interfering links. The numbers
scheduling. In the frequency scheduling inside the circles represent the node ids.
phase, the receivers, i.e. the parents on the The second part of the figure shows the tree
tree, are assigned frequencies. The goal of after time slot assignment with a single
the frequency assignment phase is to frequency channel. The numbers on the
schedule the interfering links on non- links show the assigned time slots. In this
case, it takes 6 time slots to schedule the
conicting frequencies such that the
network. In the third part of the figure we
receptions at the parents of the interfering see how the scheduling is performed with 2
senders are not disturbed. Initially, all the frequencies. First, the frequencies are
nodes operate on the same frequency. The assigned to the parents (represented inside
method collects information about the the boxes next to each parent, F1 is the
interfering links (we consider both graph initial frequency). Then, the time slots are
based and SINR based interference models assigned to the senders. With 2 frequencies,
the network is scheduled in 4 slots. The last
and present the results in Section V).
part shows the case with 3 frequencies. The
According to the collected information, at network is then scheduled in 3 time slots.
each step the most interfered parent (the We achieve a %50 reduction on the schedule
parent with the highest number of interfering length thus the data collection rate at the
links) is assigned a frequency, if one is sink node is doubled with the sufficient
available. If not, the parent node and the number of frequencies.
associated children remain on the initial The receiver based frequency assignment
makes the algorithm suitable for tree
frequency and the interference conicts have
topologies and avoids the overhead of
to be solved in the time slot assignment frequent frequency switching for the
phase. transceiver since the nodes switch at most
After the frequency scheduling, the between 2 frequencies. It may be argued
algorithm continues with the time slot that, an interfering link does not always
assignment to the senders. Similar to the disturb all the children of a parent node.
Thus, assigning a frequency to a parent
power control approach, a node can be
limits the communication possibilities for
scheduled such that it cannot transmit and those children. To investigate this issue, by
receive simultaneously and cannot receive extensive simulations, we compared the
from more than one neighbor at the same performance of our algorithm with a
time, due to the transceiver Limiations. If different approach where each sender is
the parents of all interfering senders could assigned a time slot and a frequency. The
be assigned different frequencies (this performance is observed to be similar;
however we cannot present the results in this
means the interference is totally eliminated),
paper due to the space limitations.
we can skip the SINR check. If not, during

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

sink node s (HC(s) =0), at each iteration we

D. Routing Strategies, Parent Selection add a node m T to T with the following
constraints:
In the previous sections we have discussed There is a node m Tsuch that edge (m,
the methods to cope with interference. Other m) E,
than interference, connectivity may also C (m) <m a x_degree -1,
limit the performance of scheduling. The hop count to the sink =HC (m) is
Consider the nodes that select the same minimized.
parent. They have to wait for each others
The updates are made as HC(m)= HC(m)+1
and C(m)=C(m)+1. The algorithm stops
when |T| = |V | or when no more nodes can
be added since the degree of the all nodes in
T have reached the max degree.

To clarify the gains with this method,

consider the case when all n nodes are in
range of each other and the sink. If the nodes
select their parents according the minimum
hop criteria without a degree constraint, all
the nodes will select the sink as a parent and
transmission which simply increases the
this schedule will take n time slots. On the
length of a schedule. In this section we
other hand, if we limit the number of
investigate the methods that can adjust the
connections per node as 2,t h i s will result
degree of connectivity on a tree topology.
in 2 sub trees rooted at the sink. If there is
One option would be to construct balanced
enough number of frequencies to eliminate
trees. We compared the scheduling
all the interference then the network can be
performance on minimum-hop balanced and
scheduled in 2 time slots and we achieve a
unbalanced trees. However, no improvement
factor of n/2 reduction in the schedule
is observed with balanced trees since the
length. Figure 4 shows the same network as
sink node often remains the high-degree
in Figure 3 with a different routing tree
bottleneck (due to the space limitations we
cannot represent the results in this paper). where the degree of a node is constrained to
2. The second part of the gure shows the
To avoid the bottlenecks, there should be a
time slot scheduling which takes 4 time slots
limitation on the number of children per
on a single channel frequency. The last part
parent. Thus, we explore scheduling on
shows when time slots are scheduled over
degree constrained tree topologies.
different frequency channels. This takes 2
A degree-constrained minimum-hop tree is
time slots to schedule all the links which is
constructed using a modied version of
3 times better than the baseline with a
Dijkstras shortest path algorithm. Consider
single frequency over a non-degree
a graph G(V, E) and a given degree
constrained tree, given in Figure 3.
constraint max degree. Each node n keeps a
value for the number of its children C (n)
with an initial value = 0 and hop count to the
sink HC (n) with an initial value = . The
algorithm starts with a set T that contains the

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

where is the path loss factor, Pi and Pk are

the transmission powers for i and k, and N j
is the amount of ambient noise experienced
by receiver j. Transmission is successful if
the corresponding SINR at receiver j is
4. MODELS FOR DESIGN above the threshold value:
A. Interference Models

As we discussed in the introduction, there

are two different interference models that
are commonly used in the literature:
protocol model and the physical model. Where is the SINR threshold. In Section
According to the protocol model the V-B.1, we investigate the correctness of the
transmission from a node i to a node j is graph based model and the effects of the
successful, if for every other node k both models on scheduling performance.
simultaneously transmitting, the following
condition holds B. Orthogonal Frequencies vs. Interfering
Frequencies
d(k, j) (1 + )d(i, j), ( > 0)
In the current literature on multi-channel
where d(i, j) is the distance between nodes i protocols, mostly either it is assumed that
and j, and is a guard parameter that the channels are perfectly orthogonal
ensures that concurrently transmitting nodes (interference-free) or the use of overlapping
are sufciently further away from the channels is simply avoided. Assumption of
receiver to prevent excessive interference. perfect orthogonal channels fails in practice
since radio signals are not bound to a single
On the other hand, physical model is an point in the spectrum but are distributed
SINR based model that takes into account around a mid frequency so that channel
multiple transmissions. Given a transmission overlap/interference is observed between
from node i to node j, the SINR value at adjacent bands. On the other hand, the use of
receiver j is computed as follows: only orthogonal channels cannot utilize the
spectrum efficiently. For instance, the
802.11b standards dine 11 channels of
which only three are orthogonal. Careful use
of not only 3 channels but all 11 channels by
controlling the interference can signicantly
improves the system performance [7].

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

Interference between overlapping channels We consider continuous monitoring

is inuenced by the transmission power, applications where perfect aggregation is
distance between transmitters, channel possible, i.e., each node is capable of
spacing and transceiver characteristics [8].
aggregating all the packets received from its
In Section V, we compare the impact of
orthogonal frequencies and interfering children as well as that generated by itself
frequencies on the scheduling performance into a single packet before transmitting to its
for two different transceiver platforms. parent. The size of aggregated data
Moreover, we investigate the correctness of transmitted by each node is constant and
schedules generated with the orthogonal does not depend on the size of the raw
frequencies assumption. sensor readings. Typical examples of such
aggregation functions are MIN, MAX,
5. TDMA SCHEDULING OF MEDIAN, COUNT, AVERAGE, etc.
CONVERGECASTS

In this section, we rst focus on periodic

aggregated converge cast and then on one-
shot raw-data converge cast. Our objective
is to calculate the minimum achievable
schedule lengths using an interference-aware
TDMA protocol. We rst consider the case
where the nodes communicate on the same
channel using a constant transmission
Fig. 1: Aggregated converge cast and
power, and then discuss improvements using
pipelining: (a) Schedule length of 6 in the
transmission power control and multiple
presence of interfering links. (b) Node ids
frequencies in the next section.4.1 Periodic
from which (aggregated) packets are
Aggregated Converge cast In this section,
received by their corresponding parents in
we consider the scheduling problem where
each time slot over different frames. (c)
packets are aggregated. Data aggregation is
Schedule length of 3 using BFS-
a commonly used technique in WSN that
TIMESLOTASSIGNMENT when all the
can eliminate redundancy and minimize the
interfering links are eliminated.
number of transmissions,
In Fig. 1(a) and1 (b), we illustrate the notion
Thus saving energy and improving network
of pipelining in aggregated converge cast
lifetime [19]. Aggregation can be performed
and that of a schedule length on a network
in many ways, such as by suppressing
of 6 source nodes. The solid lines represent
duplicate messages; using data compression
tree edges, and the dotted lines represent
and packet merging techniques; or taking
interfering links. The numbers beside the
advantage of the correlation in the sensor
links represent the time slots at which the
readings.
links are scheduled to transmit, and the
numbers inside the circles denote node ids.

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

The entries in the table list the nodes from signals with just enough power instead of
which packets are received by their maximum power. To this end, we evaluate
corresponding receivers in each time slot. the impact of transmission power control on
We note that at the end of frame 1, the sink fast data collection using discrete power
does not have packets from nodes 5 and 6; levels, as opposed to a continuous range
however, as the schedule is repeated, it where an unbounded improvement in the
receives aggregated packets from 2, 5, and 6 asymptotic capacity can be achieved by
in slot 2 of the next frame. Similarly, the using a non-linear power assignment [5].
sink also receives aggregated packets from We rst explain the basics of one particular
nodes 1 and 4 starting from slot 1 of frame algorithm that we use in our
2. The entries {1, 4} and {2, 5, 6} in the evaluations in Section 7.
table represent single packets comprising The algorithm proposed by El Batt et al. [6]
aggregated data from nodes1 and 4, and is a cross layer method for joint scheduling
from nodes2, 5, and 6, respectively. Thus, a and power control and it is an optimal
pipeline is established from frame 2, and the distributed algorithm to improve the
sink continues to receive aggregated packets throughput capacity of wireless networks.
from all the nodes once every 6 time slots. The goal is to nd a TDMA schedule that
Thus, the minimum schedule length is 6. can support as many transmissions as
possible in every time slot. It has two
6. IMPACT OF INTERFERENCE phases: (i) Scheduling and (ii) power control
that are executed at every time slot. First the
So far, we have focused on computing scheduling phase searches for a valid
spatial-reuse TDMA schedules where transmission schedule, i.e., largest subset of
transmissions take place on the same nodes, where no node is to transmit and
frequency at a constant transmission power. receive simultaneously, or to receive from
In this section, we focus on different multiple nodes simultaneously. Then, in the
methods to mitigate the effects of given valid schedule the power control
interference on the schedule length. First, we phase iteratively searches for an admissible
discuss the benets of using transmission schedule with power levels chosen to satisfy
power control and explain the basics of a all the interfering constraints. In each
possible algorithm. Then we discuss the iteration, the scheduler adjusts the power
advantages of using multiple channels by levels depending on the current RSSI at the
considering 3 different channel assignment receiver and the SINR threshold according
schemes. to the iterative rule: P new = SINRP
current .According to this rule, if a node
6.1 Transmission Power Control transmits with a power level higher than
what is required by the threshold value, it
In wireless networks, excessive interference should decrease its power and if it is below
can be eliminated by using transmission the threshold it should increase its
power control [6], [20], i.e., by transmitting transmission power, within the available

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

range of power levels on the radio. If all the JFTSS offers a greedy joint solution for
nodes meet the interfering constraint, the constructing a maximal schedule, such that a
algorithm proceeds with the schedule schedule is said to be maximal if it meets the
calculation for the next time slot. On the adjacency and interfering constraints, and no
other hand, if the maximum number of more links can be scheduled for concurrent
iterations is reached and there are nodes transmissions on any time slot and channel
which cannot meet the interfering constraint, without violating the constraints.
the algorithm excludes the link with Approximation bounds on JFTSS for single-
minimum SINR from the schedule and channel systems and its comparison with
restarts the iterations with the new subset of multi-channel systems are discussed in [22]
nodes. The power control phase is repeated and [23], respectively.
until an admissible transmission scenario is JFTSS schedules a network starting from the
found. link that has the highest number of packets
(load) to be transmitted. When the link loads
6.2 Multi-Channel Scheduling are equal, such as in aggregated converge
cast, the most constrained link is considered
Multi-channel communication is an efficient rst, i.e., the link for which the number of
method to eliminate interference by enabling other links violating the interfering and
concurrent transmissions over different adjacency constraints when scheduled
frequencies [21]. Although typical WSN simultaneously is the maximum. The
radios operate on a limited bandwidth, their algorithm starts with an empty schedule and
operating frequencies can be adjusted, thus rst sorts the links according to the loads or
allowing more concurrent transmissions and constraints. The most loaded or constrained
faster data delivery. Here, we consider xed- link in the rst available slot-channel pair is
bandwidth channels, which are typical of scheduled rst and added to the schedule.
WSN radios, as opposed to the possibility of All the links that have an adjacency
improving link bandwidth by consolidating constraint with the scheduled link are
frequencies. In this section, we explain three excluded from the list of the links to be
channel assignment methods that consider scheduled at a given slot. The links that do
the problem at different levels allowing us to not have an interfering constraint with the
study their pros and cons for both types of scheduled link can be scheduled in the same
converge cast. These methods consider the slot and channel whereas the links that have
channel assignment problem at different an interfering constraint should be scheduled
levels: the link level (JFTSS), node level on different channels, if possible. The
(RBCA), or cluster level (TMCP). algorithm continues to schedule the links
according to the most loaded (or most
6.2.1 Joint Frequency Time Slot constrained) metric. When no more links
Scheduling (JFTSS) can be scheduled for a given slot, the
scheduler continues with scheduling in the
next slot. Fig. 4(a) shows the same tree

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

given in Fig. 1(a) which is scheduled tree. Figure 4(b) shows the same tree given
according to JFTSS where aggregated data in Fig. 1(a) which is scheduled according to
is collected. JFTSS starts with link (2, sink) TMCP for aggregated data collection. Here,
on frequency 1 and then schedules link (4,1) the nodes on the leftmost branch is assigned
next on the rst slot on frequency 2. Then, frequency F1, second branch is assigned
links (5, 2) on frequency 1 and (1, sink) on frequency F2 and the last branch is assigned
frequency 2 are scheduled on the second slot frequency F3 and after the channel
and links (6, 2) on frequency 1 and(3, sink) assignments, time slots are assigned to the
on frequency 2 are scheduled on the last nodes with the BFS-Time Slot Assignment
slot. algorithm. The advantage of TMCP is that it
An advantage of JFTSS is that it is easy to is designed to support converge cast traffic
incorporate the physical interference model, and does not require channel switching.
however, it is hard to have a distributed However, contention inside the branches is
solution since the interference relationship not resolved since all the nodes on the same
between all the links must be known. branch communicate on the same channel.
6.2.3 Receiver-Based Channel Assignment
(RBCA)

In our previous work [7], we proposed

channel assignment method called RBCA
where we statically assigned the channels to
the receivers (parents) so as to remove as
many interfering links as possible. In
Fig. 4: Scheduling with multi-channels for RBCA, the children of a common parent
aggregated converge cast: transmit on the same channel. Every node in
(a) Schedule generated with JFTSS. (b) the tree, therefore, operates on at most two
Schedule generated with TMCP. channels, thus avoiding pair-wise, per-
(c) Schedule generated with RBCA. (b) packet, channel negotiation overheads. The
Schedule generated with RBCA. algorithm initially assigns the same channel
to all the receivers. Then, for each receiver,
6.2.2 Tree-Based Multi-Channel Protocol it creates a set of interfering parents based
(TMCP) on SINR thresholds and iteratively assigns
the next available channel starting from the
TMCP is a greedy, tree-based, multi-channel most interfered
protocol for data collection applications [8]. parent (the parent with the highest number
It partitions the network into multiple sub of interfering links). However, due to
trees and minimizes the intra tree adjacent channel overlaps, SINR values at
interference by assigning different channels the receivers may not always be high
to the nodes residing on different branches enough to tolerate interference, in which
starting from the top to the bottom of the case the channels are assigned according to

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

the ability of the transceivers to reject algorithm starts with the sink node, and adds
interference. We proved approximation a node I T at every iteration to the
factors for RBCA when used with greedy tree such that HC(i) is minimized. It stops
scheduling in [9]. Fig- ure 4(c) shows the when |T| = |V |, or when no more nodes can
same tree given in Fig. 1(a) scheduled with be added to the tree because the neighbors of
RBCA for aggregated converge cast. all these new nodes have reached the limit
Initially all nodes are on frequency F1. on their maximum degree. Consequently, in
RBCA starts with the most interfered parent, this latter situation, the heuristic might not
node 2 in this example, and assigns F2. Then always generate a spanning tree. In our
it continues to assign F3 to node 3 as the evaluation presented in Section 7.3, we
second most interfered parent. Since all consider only those instances of the
interfering parents are assigned different topologies where spanning trees with the
frequencies sink can receive on F1. specied degree constraint are produced. To
illustrate the gains of degree-constrained
7. IMPACT OF ROUTING TREES trees, consider the case when all the N nodes
are in range of each other and that of the
Besides transmission power control and sink. If the nodes select their parents
multiple channels, the network topology and according to minimum-hop without a degree
the degree of connectivity also affect the constraint, then all of them will select the
scheduling performance. In this section, we sink, and this will give a schedule length of
describe schemes to construct topologies N. However, if we limit the number of
with specic properties that help to reduce children per node to 2, then this will result in
the schedule length. two sub trees rooted at the sink, and if there
are enough frequencies to eliminate
7.1 Aggregated Data Collection interference, the network can be scheduled
using only 2 time slots, thus achieving a
We rst construct balanced trees and factor
compare their performance with unbalanced of N/2 reduction in the schedule length.
trees. We observe that in both cases the sink
often creates a high-degree bottleneck. To
overcome this, we then propose a heuristic,
as described in Algorithm 3, by modifying
Dijkstras shortest path
Algorithm to construct degree-constrained
trees Note that constructing such a degree-
constrained tree is NP-hard. Each source
node i in our heuristic keeps track of the
number of its children, C(i), which is
initialized to 0, and a hop count to the sink,
HC(i), which is initialized to. The

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

7.2 Raw Data Collection

As emphasized in [13], routing trees that

allow more parallel transmissions do not
necessarily result in small schedule lengths.
described in Algorithm 4, based on the
For instance, the schedule length is N for a
greedy scheme presented by Dai et al. [25],
network connected as a star topology,
which solves a variant of the CMST
whereas it is (2N 1) for a line topology
problem by searching for routing trees with
once interference is eliminated. Theorem 1
an equal number of nodes on each branch.
suggests that the routing tree should be
We augment their scheme with a new set of
constructed such that all the branches have a
rules and grow the tree hop by hop outwards
balanced number of nodes and the constraint
from the sink. We assume that the nodes
nk< (N +1) /2holds. In this section, we
know their minimum-hop counts to sink.
construct such routing trees.
Rule 1: Nodes with single potential parents
A balanced tree satisfying the above
are connected rst.
constraint is a variant of a capacitated
Rule 2: For nodes with multiple potential
minimal spanning tree (CMST) [24]. The
parents, we rst construct their growth sets
CMST problem, which is known to be NP-
(GS) and choose the one with the largest
complete, is to determine a minimum-hop
cardinality for further processing, breaking
spanning tree in a vertex weighted graph
ties based on the smallest id. We dene the
such that the weight of every sub tree linked
growth set of a node as the set of neighbors
to the root does not exceed a prescribed
(potential children) that are not yet
capacity. In our case, the weight
connected to the tree and have larger hop
of each link is 1, and the prescribed capacity
counts.
is (N +1) /2. Here, we propose a heuristic, as

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

Rule 3: Once a node is chosen based on the set for b1 is 5, and that for b2 is 4. So we
growth sets according to Rule 2, we attach node 4 to b2, and in the next step
construct search sets (SS)to decide which attach node 8t ob1. This balances out the
potential branch the node should be added number of nodes over the two branches.
to. A search set is thus branch-specic and
includes the nodes that are not yet connected
to the tree and are neighbors of a node that
are at a higher hop count. In particular, if the
chosen node has access to branch b, and has
a neighbor that can connect to only branch b
if b is selected, then this neighbor and its
potential children are included in the search
set for b. However, if the neighbor
has access to at least one other branch even Fig. : Balanced tree construction: Node 4 is
after b is selected, then it is not included in attached to b2
the search set. based on the search sets; load on both b1
The search sets guarantee that the choices and b2 is 5
for the nodes at longer hops to join a
particular branch are not limited by the 8. CONCLUSIONS
decision of the joining node. This balances
out the number of nodes on different In this paper, we studied fast converge cast
branches and prevents one to grow faster in WSN where nodes communicate using a
than others. Once the search sets are TDMA protocol to minimize the schedule
constructed, we choose the branch for which length. We addressed the fundamental
the sum of its load (W) and the size of the limitations due to interference and half-
search set is minimum. duplex transceivers on the nodes and
To illustrate the merit of search sets, explored techniques to overcome the same.
consider the situation. Dotted lines represent We found that while transmission power
potential communication links and solid control helps in reducing the schedule
lines represent already included tree edges. length, multiple channels are more effective.
At this point, node 4 is being processed, and We also observed that node-based (RBCA)
the loads on branches b1 and b2 are 2 and 4, and link-based (JFTSS) channel assignment
respectively, where bi denote the branch schemes are more efcient in terms of
rooted at node i. The search set SS(4,b1) is eliminating interference as compared to
{8,9,10}, because the neighbor node 8 has assigning different channels on different
access to only b1 if b1 is selected by node 4. branches of the tree (TMCP).
However, the search set SS(4,b2) is empty, Once interference is completely eliminated,
because the neighbor node 8 has access to we proved that with half-duplex radios the
another branch b1 (via node 3). Therefore, achievable schedule length is lower-bounded
the sum of the load and the size of the search by the maximum degree in the routing tree

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

for aggregated converge cast, and by max provide an order of magnitude improvement
(2nk 1,N) for raw-data converge cast. compared to the simple approach of
Using optimal converge cast scheduling scheduling on a single channel with
algorithms, we showed that the lower minimum-hop routing trees.
bounds are achievable once a suitable
routing scheme is used. Through extensive REFERENCES
simulations, we demonstrated up to an order
of magnitude reduction in the schedule [1] S. Gandham, Y. Zhang, and Q. Huang,
length for aggregated, and a 50% reduction Distributed time-optimal scheduling for
for raw-data converge cast. In future, we convergecast in wireless sensor networks,
will explore scenarios with variable amounts Computer Networks, vol. 52, no. 3, pp. 610
of data and implement and evaluate the 629, 2008.
combination of the schemes considered. [2] K. K. Chintalapudi and L. Venkatraman,
On the design of mac protocols for low-
We have explored a number of techniques to latency hard real-time discrete control
enhance the aggregated data collection over applications over 802.15.4 hardware, in
a tree topology in WSN. Our initial IPSN 08, pp. 356367.
approach was to use interference-aware [3] I. Talzi, A. Hasler, S. Gruber, and C.
minimum frame-length TDMA-scheduling Tschudin, Permasense: investigating
that enables spatial reuse. The second step permafrost with a wsn in the swiss alps, in
was to combine the scheduling with EmNets 07, Cork, Ireland, pp. 812.
transmission power control. Although the [4] S. Upadhyayula and S. Gupta, Spanning
well studied transmission power control tree based algorithms for low latency and
method helped to overcome interference and energy efcient data aggregation enhanced
reduce schedule length, it was found to be convergecast (dac) in wireless sensor
not always the best solution in a practical networks, Ad Hoc Networks, vol. 5, no. 5,
setting due to the limitations on the power pp. 626648, 2007.
settings of the nodes. The next step was to [5] T. Moscibroda, The worst-case capacity
consider the use of multiple frequency of wireless sensor networks, in IPSN 07,
channels. With the extensive simulations we Cambridge, MA, USA, pp. 110.
found that for networks of about a hundred [6] T. ElBatt and A. Ephremides, Joint
nodes, the use scheduling and power control for wireless
of multi-frequency scheduling can suffice to ad-hoc networks, in INFOCOM 02, Jun,
eliminate most of the interference. Then, pp. 976984.
data collection rate was no longer [7] O. D. Incel and B. Krishnamachari,
interference-limited, but rather topology- Enhancing the data collection rate of tree-
limited. Thus, our nal approach was to use based aggregation in wireless sensor
an appropriate degree-constrained tree networks, in SECON 08, San Francisco,
construction. Simulation results showed that, CA, USA, pp. 569577.
combining the last two techniques can

[email protected] all rights reserved www.ijrct.org

 V Redya Jadav/International Journal of Research and Computational Technology, Vol.5 Issue.2
ISSN: 0975-5665, May, 2013 www.ijrct.org

[8] Y. Wu, J. Stankovic, T. He, and S. Lin, TECHNOLOGICAL AND RESEARCH

Realistic and efcient multi-channel INSTITUTE from Jawaharlal Nehr
communications in wireless sensor u Technological University, Hyderabad, AP
networks, in INFOCOM 08, pp. 1193 and India.
1201.
[9] A. Ghosh, O. D. Incel, V. A. Kumar, and
B. Krishnamachari, Multi-channel
scheduling algorithms for fast aggregated
con- vergecast in sensor networks, in
MASS 09, Macau, China.
A.Devilal received the
B.Tech, degrees in computer science and
engineering from Jawaharlal Nehru
Technological University, Hyderabad, AP,
and India. He is Pursuing M.Tech in
Computer Science and Engineering at
V.Redya received BOMMA INSTITUTE OF TECHNOLOGY
the B.E, degrees in computer science and and SCIENCE from Jawaharlal Nehru
engineering from Osmania University, Technological University, Hyderabad, AP
Hyderabad, AP, and India. He is received and India.
M.Tech in Computer Science and
Engineering at Jawaharlal Nehru
Technological University, Hyderabad, AP
and India. He is pursuing Ph.D program in
computer science and Engineering.

D.Sravan received the

B.Tech, degrees in computer science and
engineering from Jawaharlal Nehru
Technological University, Hyderabad, AP,
and India. He is received M.Tech in WEB
TECHNOLOGIES at AURORAS

[email protected] all rights reserved www.ijrct.org