Quality of Service On-Demand Power Aware Routing Algorithm For Wireless

Sensor Networks
Deepali Virmani1, Satbir Jain2
B5/107 Mayur Apartment , Sector 9 , Rohini, Delhi , India.

deepalivirmani@gmail .com

Abstract
Wireless Sensor networks are non-infrastructure networks which consist of mobile sensor nodes. Since the mobile nodes have
limited battery power, it is very important to use energy efficiently in sensor networks. In order to maximize the lifetime of
sensor networks, traffic should be sent via a route that can avoid nodes with low energy thereby minimizing the total
transmission power. In addition, considering that the nodes of sensor networks are mobile, on-demand routing protocols are
preferred for sensor networks. In this paper, we propose a novel on-demand power aware routing algorithm supporting quality of
service called QDPRA. QDPRA prolongs its network lifetime by compromising between minimum energy consumption and fair
energy consumption without additional control packets. QDPRA also improves its data packet delivery ratio, minimizes delay
and maximizes throughput of the network .

Keywords : SPIN, LEACH, DDiff, RTREQ, QDPRA.

1 Introduction
A sensor network is a collection of wireless sensor nodes that come together to form a self-organizing network
without any support from the existing fixed communication infrastructure. In such a network, each device plays the
role of a router and has limited battery energy. Thus, it is widely accepted that conventional routing protocols are not
appropriate for sensor networks and consequently, the design of routing protocols for such networks is a challenging
issue taking power factor into consideration. To reduce the energy consumption in mobile devices, there have been
efforts in physical and data link layers as well as in the network layer related to the routing protocol. The physical
layer can save energy by adapting transmission power according to the distance between nodes. At the data link
layer, energy conservation can be achieved by sleep mode operation. The purpose of on demand power-aware
routing protocols is to maximize the network Lifetime and minimize energy consumption by only transmitting
message when there is demand of that message( Sleep and Wake technology ). The network lifetime is defined as
the time when a node runs out of its own battery power for the first time [1]. If a node stops its operation, it can
result in network partitioning and interrupt communication. The on demand power-aware routing protocols should
consider energy consumption from the viewpoints of the network , sleep mode and the node levels. From the
network point of view, the best route is one that minimizes the total transmission power. On the other hand, from the
viewpoint of a node, it is one that avoids the nodes with lower power. And from the point of sleep mode ,it should
only operate when demanded otherwise node should be in sleep mode. It is difficult to achieve these all the
objectives simultaneously. Minimizing the total energy consumption tends to favor min-hop routes. However, if the
min-hop routes repeatedly include the same node, the node will exhaust its energy much earlier than the other nodes
and the network lifetime will decrease.
This paper proposes an on-demand power-aware routing algorithm supporting quality of service called
QDPRA (Quality of Service On-Demand Power Aware Routing Algorithm). Our proposed routing algorithm
balances between minimum transmission energy consumption and fair node energy consumption in a distributed
manner. This goal is achieved by controlling the rebroadcast time of RTREQ packets. In addition, we design a
mechanism of estimating the average energy level of the entire network without additional control packets. The
estimated average energy is useful to adaptively control the rebroadcast time.
The rest of the paper is organized as follows. Section 2 reviews typical routing algorithms and discusses the
pros and cons from the viewpoint of the network lifetime. In Section 3, we present our proposed power-aware
routing algorithm in detail. Section 4 describes the simulation results and performance comparison. Finally, we
conclude this paper in Section 5.

2 ) Overview of Routing Algorithms

2.1) Sensor Protocols for Information via Negotiation (SPIN): Heinzelman et.al. in [2] [3] proposed a family of
adaptive protocols called Sensor Protocols for Information via Negotiation (SPIN) that disseminate all the
information at each node to every node in the network assuming that all nodes in the network are potential base-
stations. This enables a user to query any node and get the required information immediately. These protocols make
use of the property that nodes in close proximity have similar data, and hence there is a need to only distribute the
data that other nodes do not posses. One of the advantages of SPIN is that topological changes are localized since
each node needs to know only its single-hop neighbors. SPIN provides much energy savings than flooding and
meta- data negotiation almost halves the redundant data. Using SPIN routing algorithm, sensor nodes can conserve
energy by sending the metadata that describes the sensor data instead of sending all the data. SPIN can reduce the
power consumption of individual node, but it may decrease the lifetime of the whole network due to extra messages.

2.2) Low-Energy Adaptive Clustering Hierarchy (LEACH): is a clustering-based protocol that minimizes energy
dissipation in sensor networks [4]. LEACH randomly selects sensor nodes as cluster heads, so the high energy
dissipation in communicating with the base station is spread to all sensor nodes in the sensor network. However,
data collection is centralized and is performed periodically. Therefore, this protocol is most appropriate when there
is a need for constant monitoring by the sensor network. LEACH can suffer from the clustering overhead, which
may result in extra power depletion.

2.3) Directed Diffusion( DDiff): In [5], C. Intanagonwiwat et. al. proposed a popular data aggregation paradigm for
WSNs, called directed diffusion. Directed diffusion is a data-centric (DC) and application- aware paradigm in the
sense that all data generated by sensor nodes is named by attribute-value pairs. The main idea of the DC paradigm is
to combine the data coming from different sources and combines them by eliminating redundancy, minimizing the
number of transmissions; thus saving network energy and prolonging its lifetime. But still power consumption is
high .
Therefore we propose an algorithm that tries to prolong the network lifetime by compromising between
minimum energy consumption and fair energy consumption without additional control packets. QDPRA also
improves its data packet delivery ratio, minimizes delay and maximizes throughput of the network.

3) Quality of Service On-Demand Power Aware Routing Algorithm (QDPRA)

3.1) Basic Idea
Generally in on-demand routing protocols [6][7], the source floods an RTREQ (Route-Request) packet to search a
path from source to destination. The destination node receives the RTREQ packet and unicasts an RTREP (Route-
reply) packet to the source to set up a path. Likewise, our proposed QDPRA is an on-demand algorithm. QDPRA
doesn’t use additional control packets to acquire necessary information for power aware routing but utilizes RTREQ
packets which are already used in on-demand routing protocols. QDPRA only requires the average residual battery
level of the entire network, which can be obtained without any control packets other than RTREQ packets. In our
proposed algorithm, intermediate nodes control the rebroadcast time of the RTREQ packet, where retransmission
time is proportional to the ratio of average residual battery power of the entire network to its own residual battery
power. In other words, nodes with relatively larger battery energy will rebroadcast RTREQ packets earlier. Because
on-demand routing protocols drop duplicate RTREQ packets without rebroadcasting them, QDPRA can set up the
route composed of the nodes with relatively high battery power.

3.2) Average Residual Battery Power Estimation

Basically the nodes use their residual battery power for the rebroadcast time of RTREQ packets. If the time is
determined only by the nodes absolute residual battery power, then the retransmission time will increase as time
passes by. Therefore, the relative measure should be used. As a relative measure, we used the average residual batter
power of the entire network. The exact value of this average power can be acquired by periodic control packets, but
using periodic control packets isn’t an on-demand method and it also consumes more energy.

To estimate the average energy, our proposed algorithm uses only RTREQ packets that are already used in on-
demand routing. For this end, AR and NR fields are added to the packet header, where AR is the average residual
battery power of the nodes on the path and NR is the number of hops that the RTREQ packet has passed. The
mechanism to obtain the estimated average value is as follows.

1. First, the source records its own battery power to the AR field, and sets the NR to 1, and broadcasts the
RTREQ packet.
 2. Assume that a node x has received an RTREQ packet, and the node x’s residual batter power is BPx and
the AR value of the RTREQ packet is ARold. Then the average residual battery power, ARnew, of new route
that includes the node x is as following
AR old × NR + BP x
AR new = (1)
NR + 1
Before the node x rebroadcasts the packet, it updates AR to ARnew and increases the value of NR by one.
This step is not executed for duplicate RTREQ packets.

3. Whenever a node x receives an RTREQ packet, it calculates the average residual battery power of the
network by the following equation.
AEr new = (1 − β ) AEr old + β AR old (2)
where β is the weighting factor of the moving average. The β is set to 0.75 in our simulations.

3.3) Rebroadcast Time Control

A node x determines its rebroadcast time TR as follows.
 AEr 
TR = DT × 
 BP  (3)
 x 
is the estimated average power, BPx is its own residual power, and DT is a constant to scale the retransmission time.
According to equation (3), if the residual battery power Bpx is smaller than the average network residual power AEr,
then the retransmission time DT will be longer, and if BPx is larger then vice versa. So if the individual battery power
BPx is larger than the average, then the node x would tend to be selected as a member of the route, which results in
fair energy consumption among the nodes. When the residual battery power variation is small, most nodes have a
similar retransmission time. In that cast, the route with a smaller hop count will be selected. This shows that
QDPRA compromises between the min-hop path and the fair energy consumption path.

3.4) Delay Metric

Each node includes in the Hello message, during the neighbor discovery performed by the QDPRA, the creation
time of this message. When a neighbor node receives this message, it calculates the difference between such a time
and the current time, which represents the MD (Measured Delay) .This delay, includes the queuing time, the
transmission time, the collision avoidance time and the control overhead time. Otherwise, the measurement of the
one-way delay avoids the increase of traffic load at adding acknowledgment messages to the QDPRA protocol.
Average Delay (AD) is calculated as follows.
AD = γ × AD + (1 − γ ) MD (4)

Where γ is the delay parameter and γ is set to 0.5 in our simulations.

3.5) Bandwidth metric

A flow is formed by the packets with the same pair of source and destination. When a data flow takes a path P, it
interacts with other flows in the near neighborhood of the nodes which compose P. The principle in a wired network
is simpler because flows interact only between each other, when they go through the same link. Moreover, nodes can
reserve bandwidth. The higher bandwidth used by flows is the higher performance of the algorithm. The saturate
bandwidth Bsat [8] is the minimum bandwidth on a link when all the flows of this link have the same bandwidth. It
can be represented on a graph (G, V) by the function:
bx , y +1 
∀( Ax , Ax +1 ) ∈V Bsat = min   (5)
 Nf x 
Where bx , y +1 is the bandwidth of link ( Ax , Ax+1 ) and Nf=x is the number of flows on this link. Highest saturate
bandwidth can be calculated by the formula :

(
Bsat max = max Bsat R1 ,  , Bsat Rn ) (6)
Where R1 ,…….,Rn are all possible loop free routing links for a given network and BsatRz is the saturate bandwidth
yielded by the algorithm. The algorithm which returns B satmax is optimal . So the problem is to find the saturate
bandwidth.
In wireless network, collisions may interfere with packet transmissions. Moreover, it is difficult to reserve a part of
bandwidth. Many algorithms use the TDMA access method to reserve a bandwidth part on a shared channel.
However, nodes need to be synchronized [9], which is tough to realize in MANETs. We modify the definition of the
saturate bandwidth so that the network collisions are taken into account. As a consequence, the saturate bandwidth
can be represented in a graph (G, V) by:
 min ( SPx , SPx +1 ) − Bcoll 
∀( Ax , Ax +1 ) ∈V Bsat = min   (7)
 Nf x 
Where SPx is the speed of the node interface x and Nfx is the number of flow of the link. (x , x+1 ) . Bcoll is the
bandwidth with collision on this link. is necessary to calculate the number of packets transmitted by the link layer. In
[10] a method to calculate the time for a collision-free packet transmission is given. So the bandwidth seen by one
packet of M bits can be calculated as:
M
B packet = RT
(TTT + TCAT + TCDT + TCOT ) × RT + ∑TBT n
(8)
n =1

Where TTT is the packet transmission time, TCAT is the Collision Avoidance phase time, TCDT is the collision detection
time, TCOT is the Control Overhead time (e.g. RTS, CTS, etc.), RT the number of necessary transmissions, TTBn is the
Backoff time for nth retransmission. Then we derive the bandwidth with collisions for each packet P as follows:
MP
Bcoll ( P ) = RT
if RT P > 1
(TTT + TCAT + TCDT + TCOT ) × ( RT p − 1 ) + ∑TBT n
n =1
(9)
Bcoll ( P ) = 0 if RT p =1 (10)
The increase of the saturate bandwidth involves an increase of the flow path bandwidth. The only increase of the
bandwidth does not imply all the time an increase of the saturate bandwidth. so we use a weight function in our
proposed routing algorithm to give a path with the highest bandwidth. Finally , for a path p = (v1, ……., vn) the
weight function is :
n =1
x
∑B
x =1 x , y +1
W ( p) = (11)
B 2
B
n

1 − max  coll ,  , coll 

 B1, 2 Bn −1, n 
Our weight function is asymmetric i.e. for two paths P and P’ with w(P) ≤ w(P’), a path P’’ can exist and breaks the
order relation so w(P’,P’’) ≤ w(P,P’’).

4) Simulations

We used J-Sim ( Java simulator supporting wireless networks ) to compare the performance of QDPRA with that
of existing power-aware routing algorithms SPIN, LEACH, DDiff. As mentioned before, on-demand routing
protocols are adequate for sensor network environments. Therefore we performed simulation on the power-aware
routing algorithms that could be implemented to on-demand routing protocols.

4.1) Simulation Model

In all simulations, the same topology is used, where 100 nodes are uniformly distributed and nodes are 150m apart.
The initial energy of all the nodes is 0.3J. The transmission power is 250mW and the receiving power is 150mW. A
node waits a maximum pause time of 500 seconds. The traffic is CBR. The data packet length is 500 bytes. Each
node pauses for 70 seconds and moves to a random position at the maximum speed of 2m/s (average 1m/s).

4.2) Results
4.2.1) Network Lifetime : Figure 1 shows the number of nodes that run out of their battery power as a function of
time. The time when the first node dies indicates the lifetime of the network, and the slope of the graph shows the
fairness of energy consumption among nodes. If the slope is small, it means that the variation of the lifetime of the
nodes is large. That is, the use of batteries is unfair. On the contrary, if the slope is steep, it means that the battery
power of the nodes has been fairly used.
60

50
Total number of dead nodes

40
SPIN
LEACH
30
DDiff
QDPRA
20

10

0
3

9
51

57

65

82
40

45

73

99

14

17
12

18

Fig 1 Expiration
Time (secs)
Sequence of nodes
Comparing the network lifetime of each algorithm, the SPIN had the shortest lifetime of 52 seconds. SPIN routing
also had the smallest slope, which means the energy was consumed unfairly among the nodes. Although LEACH
extended the network lifetime to approximately 78 seconds by using battery power evenly among the nodes, the
lifetime extension wasn’t so good since it tended to select long paths with many hops to guarantee fairness. DDiff
minimized the network energy consumption and extended its network lifetime to 93 seconds. As DDiff used
additional control packets, the network lifetime didn’t increase dramatically compared to SPIN. QDPRA showed
better performance than the others. The network lifetime increased to 128 seconds which is about 2.5 times longer
than that of SPIN routing. The network lifetime is about 1.4 times longer than that of DDiff. This improvement is
due to the fair energy consumption without additional control packets.

4.2.1) Delivery Ratio: Figure 2 shows a comparison of the delivery ratio among power-aware routing algorithms.
We can see that the better power-aware routing algorithms also have a better delivery ratio. QDPRA showed the
highest delivery ratio of about 95% with the increasing number of nodes, which is approximately 13% higher than
that of SPIN and approximately 7% higher than that of DDiff. The delivery ratio is increased because residual
battery power are excluded from the route in power-aware routing algorithms.

100

90

80

70
% Delivery ratio

60 SPIN
LEACH
50
DDiff
40 QDPRA

30

20

10

0
10 20 30 40 50 60 70 80 90 100
Fig 2 % Delivery Ratio
No of nodes
4.2.3) Saturate Bandwidth : In figure 3, we show the saturate bandwidth according to the number of flows in
network of 10 nodes. The saturate bandwidth decreases with the increase of the number of flows, because when the
number of flows grows the saturate link is crossed by a higher number of flows. As compared with previous
algorithms our proposed algorithm shows most promising results but maximizing the bandwidth utilization.

800

700

Saturate Bandwidth(Kb/s)
600

500
SPIN
LEACH
400
DDiff
QDPRA
300

200

100

0
10 20 30 40 50 60 70 80 90 100
Fig 3 Saturate Bandwidth
No of Flows

4.2.4) Packets With Collision: The analysis of simulation traces in figure 4 shows that our algorithm chooses paths
with few collisions than those chosen by SPIN, LEACH and DDiff. The reason is that the path doesn’t change,
except when a link is down.
1200
NO of Packets with Collision

1000

800
SPIN
LEACH
600
DDiff
QDPRA
400

200

0
10 20 30 40 50 60 70 80 90 100
Fig 4 Packets with Collision
No of Flows

4.2.5) Average Delay: Figures 5 depicts an improvement of average transmission time i.e. the complete time
between a data packet leaves the sender node and arrives in the destination node. As expected, the average
transmission time is reduced in QDPRA , minimizing the average delay and a gain of 17% is achieved.
600

500
Average Transmission Time(ms)

400
S PIN
LE ACH
300
DDiff
QDPRA
200

100

0
10
Fig205 Average
30 40 50 60 70
Transmission 80
Time90 100
No of Flow s
5) Conclusion
In this paper, we proposed a new power-aware routing algorithm called QDPRA. QDPRA is an on-demand routing
protocol which sets its route in a distributed manner. QDPRA only requires average residual battery level of the
entire network, which can be obtained without other control packets except for RTREQ packets. When RTREQ
packets are broadcasted, the rebroadcast time is determined by the amount of time which is proportional to the ratio
of average residual battery power of the entire network to its own residual battery power. As a result, QDPRA
selects nodes that have relatively abundant battery energy. Since the rebroadcast time dynamically varies according
to residual battery power, QDPRA keeps a balance between min-hop routing and fair battery consumption. The
simulation results showed several advantages of QDPRA over other existing algorithms in terms of performance.
QDPRA not only prolongs the network lifetime but also improves the delivery ratio by selecting a more reliable
path. As well as QDPRA maximizes the bandwidth utilization and minimizes the delay by avoiding routes having
collision.

References

1. Q. Li, J. Aslam and D. Rus: Online Power-aware Routing in Wireless Ad-Hoc Networks. In: MOBICOM, pp 303—
309, Newyork (2001).

2. W. Heinzelman, J. Kulik, and H. Balakrishnan, :Adaptive Protocols for Information Dissemination in Wireless
Sensor Networks. In: . 5th ACM/IEEE Mobicom Conference (MobiCom '99), pp. 174 – 185, Seattle, WA,
August, (1999).

3. J. Kulik, W. R. Heinzelman, and H. Balakrishnan, : Negotiation-based protocols for disseminating information in

wireless sensor networks. Wireless Networks, Volume: 8, pp. 169-185,(2002).

4. L. Subramanian and R. H. Katz, :An Architecture for Building Self Configurable Systems.In: IEEE/ACM Workshop
on Mobile Ad Hoc Networking and Computing, Boston,pp 45 – 51, MA, August (2005).

5. C. Intanagonwiwat, R. Govindan, and D. Estrin, : Directed diffusion: a scalable and robust communication paradigm
for sensor networks. In: ACM MobiCom '05, pp. 56 -- 67 Boston, MA, (2000).

6. D. Johnson, D. Maltz: Dynamic Source Routing Ad Hoc Wireless Networks. Mobile Computing, Kluwer Academic
Publishers (1996).

7. C. Perkins, E. Royer: Ad-hoc On-Demand Distance Vector Routing. In: 2nd IEEE Workshop on Mobile Computing
Systems and Applications, pp 78 – 85, (1999).

8. J. Wang and K. Nahrstedt, : Hop-by-Hop Routing Algorithm for Premium-class Traffic in Diffuser Networks,
In : INFOCOM 2002, pp 23 –39,(2002).

9. J. Elson and K. Romer, : Wireless Sensor Networks: A New Regime for Time Synchronization, In: ACM Computer
Communication Review (CCR), Vol. 33 No. 1, pp. 149-154, (2003).

10. M. Kazantzidis and M. Gerla, : End-to-end versus Explicit Feedback Measurement in 802.11 Networks, In : 7th
IEEE Symposium on Computers and Communications, 2002.