A Reactive Power-Aware on-Demand Routing
Protocols for Wireless Ad Hoc Networks
Chao Gao, Riku J¨ ntti
Department of Computer Science
University of Vaasa
Vaasa, Finland, FIN-65101
Email: gc@puv.ﬁ, riku@uwasa.ﬁ
Abstract— Ad hoc network routing protocols can be designed the number of stations, and the network grid area. Based on
to give better energy conservation performance. In this paper, these conclusions, they proposed two power control schemes:
we summarize the problems of power control in wireless ad Common Power Control (CPC) for entire network scale and
hoc networks. Main discussion is to propose an energy-efﬁcient
reactive routing protocol. For such a routing protocol some Independent Power Control (IPC) for hop-by-hop scale. In 
related topics are considered, such as MAC layer requirement, and , power control for energy efﬁcient routing protocols
transmitter power imbalance, route-establishment, and route was proposed. In , it also gives the comparison of energy
stability. Our protocol can be applied to an existing reactive efﬁciency between multi-hop link and single hop link. The
routing protocol such as AODV or DSR to give signiﬁcant efﬁcient authors declared that an intermediate node acting as relaying
power conservation performance. A simulation model is used to
test the performance and the results show that our protocol gives device will decrease the energy consumption if its location is
signiﬁcant power saving and improves link stability as well. good.
In this paper we are going to analysis the ad hoc network
topology properties and propose a new energy efﬁcient routing
I. I NTRODUCTION protocol which can be adapted to the existing reactive routing
Wireless ad hoc networking has gained quite much attention algorithms such as AODV  and DSR .
in recent years. Typically an ad hoc network is a peer-to-peer The rest content of this paper is arranged so: Section II will
radio network without any infrastructure. While the problems illustrate transmitter-receiver power control scheme which is
in the physical layer (such as radio interface, modulation) and called hop-by-hop power control. Section III will consider the
the data link layer (such as multiple access control - MAC) are power control at the network layer, that is, power coordination
heavily discussed, the network layer problems are considered among the nodes in a network. We also propose a routing
to be very important as well. However, a key issue in ad hoc protocol in this section based on the previous discussion. In
networks is energy conservation, because nodes or devices in Section IV, we will give simulation results and analysis of our
such a network are usually mobile and battery operated. Either routing scheme from energy efﬁciency point of view.
physical layer, or data link layer, or network layer protocols
may contribute to energy conservation. While work in the
two lower layers is still progressing, network layer energy II. H OP - BY-H OP P OWER C ONTROL
conservation is thought to be another most promising point Hop-by-hop power control is proposed in  and . This
under the current technology level. method requires that each node can put the transmit power
Among these approaches, power control is an effective level PT X , at a suitable format ﬁeld in the transmitted packet.
method for energy efﬁciency. Power control is a part of digital It also requires that the radio receiver can measure the received
cellular networks such as TDMA and CDMA and has very signal strength, PRX . With these two values, the node that
sophisticated implementation in these networks. However, in has received the packet can estimate the link attenuation.
ad hoc networks because there is no base station working In another word it can estimate the distance between the
as coordinator and connection may occur between any two transmitter node and itself. Upon the received signal power,
nodes in the network, power control is a much more complex the node can adjust its transmission power to the remote node
problem than that in cellular systems. In many literatures this by:
issue has been mentioned recently. In , power control as a PT X = PT X − PRX + SR + M
method is demonstrated to be beneﬁt to increase trafﬁc capac-
ity of entire network, increase battery lifetime, and reduce the Here SR is the minimum signal power required to correctly
contention at the MAC layer. It further gives a power-control- receive a packet. M is a power safety margin introduced to
enabled routing protocol for proactive routing schemes such as take into account channel and interference power level ﬂuctu-
DSDV. In  and , power control implementation on MAC ations. When a packet from this node is sent and received by
layer has been proposed. The authors in  asserted that the the remote node, the remote node can do the same adjustment.
optimal transmission power is determined by the network load, This will result to a close loop power control.
Coverage of A
Coverage of D Coverage of
Fig. 1. Power Imbalance Problem
Here arises a so-called power imbalance problem. Let’s
suppose the ad hoc network here we are talking about follows
IEEE802.11 MAC protocol , which uses RTS/CTS hand- Fig. 2. Simultaneous Transmissions between A and B, C and D are possible,
shake to avoid collisions. If two nodes are transmitting a packet but prohibited by RTS/CTS
to their own destinations at the same time, one destination node
may be interfered by another transmitting node which is not
aware of this problem. This is illustrated in Fig.1: When node transceiver, i.e. with the same maximum transmit power and
C has a packet to D, it broadcasts RTS packet and D replies the same variation scale.
a CTS packet by adjusted power. This signal cannot be heard
by A, because the distance between C and D is shorter than A. Uplink and Downlink Power Control
that between A and B. Thus both A and C may send a packet In AODV or DSR routing protocol, when a route is being
simultaneously and cause collision at D. established, the source node initiates a Route Request (RREQ)
This problem can be solved by setting maximum power packet, which is broadcast to all its neighbours. For power
when transmitting RTS and CTS packets. That is, in Fig.1, control issue, we assign each intermediate node two distin-
when node D has received a RTS packet (which is sent with guish transmit power values for uplink PT Xup and downlink
maximum power), it replies a CTS packet with maximum PT Xdown , respectively. Of course, for the source and the
power. This will of course prohibit any possible transmission destination nodes, there is only one direction thus either PT Xup
in its full power coverage. Node A will be notiﬁed and will or PT Xdown is available. As depicted in Section II, when a
not send anything during the Network Allocation Vector (NAV) node sends a packet, it embeds the transmitter power into
time speciﬁed by the CTS packet from D. However, when we the packet, which is RREQ at here. Because the source node
consider the situation given in Fig.2, this solution will prohibit doesn’t know the location of its neighbours, it will broadcast
the simultaneous transmission even they may not interfere each the RREQ packet with maximum power. The nodes that have
other. received the RREQ can adjust its uplink transmit power by
The same broadcast will continue by the neighbours with
III. POWER CONTROL ROUTING maximum power until the RREQ packet reaches the destina-
So far, the routing algorithms applied in ad hoc networks tion. The destination node will send a Route Reply (RREP)
can be classiﬁed into two categories: proactive and reactive. packet to its uplink node, the same, with embedded transmit
Proactive routing is based on the establishment of routing table power value. The node that receives this RREP can now adjust
which is periodically updated by exchanging control packets its downlink transmit power by Equ.(1).
throughout all the nodes. In reactive routing protocols, route This procedure can be illustrated by Fig.3.
discovery procedures are invoked on demand when a source After the route establishment procedure, Equ.(1) can be
has a new connection pending a new destination. In , deployed for both uplink and downlink power adjustment
performance comparison was inspected between proactive and along the route when DATA packets and their ACK are
reactive routing protocols with the respect of energy efﬁciency. exchanged hop-by-hop.
The result shows reactive routing protocols such as AODV and
DSR give better performance then their proactive counterparts. B. Energy Efﬁcient Route Finding
Here we propose a power control routing scheme for reactive When a RREQ packet is sent with maximum power as
routing protocols. depicted previously, it is possible that several neighbours with
We assume that a route in ad hoc networks is always a different distance to the source will receive it. This may result
duplex link. All the nodes in the network have normalized the to that a less efﬁcient route will be selected, as shown in Fig.4.
In order to ﬁnd the optimal energy efﬁcient route, two rules
are deployed. The ﬁrst is every intermediate node holds a
1. Initiates RREQ
with Pmax transmission back-off interval according to the received RREQ
S 1 D signal power. As mentioned at beginning, we’ve assumed that
2. Get uplink Ptxup all the nodes here have normalized transmitter power. Upon
3. Forwards RREQ
with Pmax receiving a RREQ packet, it is possible for the node to estimate
the distance between the uplink node and itself by radio
(a) RREQ Propagation
PRX = (2)
According to , there exists an optimal distance Dopt
4. Get uplink Ptxup
8. Get downlink 5. Initiates RREP
by which the minimum hop-by-hop power consumption is
Ptxdown with Ptxup achieved. We denote the received signal power as PRXopt
S 1 D
when Dopt is given by Equ.(2).
6. Get downlink Ptxdown
7. Forwards RREP
The back-off time is set according to the ratio of actual
received signal power to PRXopt , denoted as:
Tbk = c PRX − PRXopt (3)
(b) RREP loopback
Here c is a constant. This rule will guarantee that a better
hop will be chosen. As in Fig.4, node 1 will send RREQ before
Fig. 3. Route Establishment with Power Control
node 2 for sure.
The second rule is about multiple RREQ receptions. If we
look at Fig.4 again, we will get node 1 to relay the RREQ ﬁrst.
Both nodes 2 and 3 will receive it, but node 2 has received
one RREQ already. By original rule node 2 will discard the
second RREQ. Unfortunately this will result to route S-1-3-
D selected. However, it is possible for node 2 to compare
S 1 2 3 D
the received signal power of these two RREQ packets. If the
PRX1 − PRXopt > ∆ + PRX2 − PRXopt (4)
then the node that receives the new RREQ will replace the
Optimal Route uplink node by the current one. Here PRX1 and PRX2 are
Less Effective Route the received signal power of previous RREQ and current
RREQ, respectively. ∆ is a safety margin to tolerate signal
Fig. 4. Route Selection with Maximum RREQ Power ﬂuctuations. It means the node that generates the second
RREQ is closer to the optimal distance. The node will replace
the previous uplink node by the current one. Back to Fig.4, if
In Fig.4, a route is to be established from node S to node D. node 2 backs-off its RREQ, it will receive the second RREQ
According to , a moderate hop distance will perform best from node 1, with the signal power satisfying Equ.(4). Thus
energy efﬁciency; therefore the optimal energy efﬁcient route node 2 will replace the uplink node from S to 1.
is S-1-2-3-D. However, S initiates a RREQ with maximum Because the second RREQ that node 2 received has better
power; both nodes 1 and 2 will receive it. First there is a power condition to Equ.(3), node 2 will send the RREQ before
contention between node 1 and 2 to relay the RREQ. If node that of node 3, which has received a RREQ from node 1
2 gets the chance ﬁrst, both nodes 3 and D will receive the already. The same update will happen on node 3 thus the
RREQ and D will send RREP upon the reception immediately. optimal energy efﬁcient route S-1-2-3-D is established.
Thus a less efﬁcient route S-2-D with two hops is established. One situation needs to be considered, as shown in Fig.5.
Actually this is true when original AODV protocol is applied. In the ﬁgure, both nodes 1 and 2 are located far away from
On the other hand, if node 1 wins the contention and forwards the source with about the same distance, so they will back-off
the RREQ packet, both nodes 2 and 3 will receive it. In the the relaying of RREQ by approximately the same Tbk . If we
case node 2 has already received the same RREQ once. It will assume that node 1 wins the contention, it will send RREQ
discard it according to the original AODV or DSR routing to 2 and 3. Node 2 will replace its uplink node from S to 1
protocol. Now it is for nodes 2 and 3 to content the channel. because the later one has better energy performance according
We will get another less efﬁcient route S-1-3-D if node 3 wins to equ.(4). When node 2 broadcasts the RREQ, node 3 will
However, the scheme can select an energy-optimal route based
on any reactive routing protocol. Meanwhile, a moderated
transmitter power will reduce the interference to other nodes
1 thus will improve the reception quality. By this means it
S increases the network throughput because the rate of retrans-
2 mission is reduced. A back-off of RREQ packet forwarding is
introduced, proportional to the gap between received signal
power and the optimal one. This will increase the route
establishment latency, however, when the node density is high,
this is not a problem.
Fig. 5. Less Efﬁcient Route The route latency in a high density network becomes lower
because of the introduction of the back-off delay Tbk that
occurs at each node to relay RREQ. In original AODV, all
Node2 receives another RREQ
the nodes in the radio coverage of a transmitting node will
after Tbk, rejects it.
receive a RREQ and all will start contention. In our scheme,
this is scattered by the distance equation (3). If the density
RREQ RREQ of network is properly high, there are always nodes located
Tp Tbk RREQ Node3 receives another RREQ
before Tbk, accepts it.
at the proper distance from the RREQ-source node and these
TX Tp Tbk
nodes will start contention to relay RREQ earlier than other
Time nodes. It means fewer nodes in the contention thus latency is
Fig. 6. Contention window plays role to select route Our scheme doesn’t increase network overhead because
each node broadcasts RREQ only once.
replace its routing data because node 2 shows a better energy
budget than that of node 1. Thus a less efﬁcient route S-1-2- A. Simulation Environment
3-D is established. A simulation model is setup for the scheme described above.
This problem can be recovered by setting a simple rule: if Different network scales are inspected by 20 nodes, 40 nodes,
the back-off interval Tbk is expired, the node will drop any 60 nodes, and 80 nodes with random distribution in an area of
duplicated RREQ. We can see that the optimal route S-2-3-D 1000x200, 1000x400, 1000x600, and 1000x800, respectively.
will be selected as illustrated in Fig.6. It means that we keep the node desity at 10000m2/node. By
this node density it is easy to check the established route with
C. Summary of Power Control Routing Protocol
two different methods: the original AODV and the power-
Here we summarize our power control routing scheme: aware routing protocol proposed in this paper. Each scenario
1) Hop-by-hop power control is achieved by Equ.(1). has been simulated 10 times and the average results are given
2) At MAC layer, RTS/CTS handshake uses maximum in the following tables. Nodes are randomly moving or pausing
power to avoid collisions. in the area and moving speed is set between SPEEDMIN to
3) RREQ is broadcast with maximum power. SPEEDMAX m/sec. After the arrival of a destination, a node
4) An intermediate node backs-off RREQ packet. The will stay there (pause) for a random time from PAUSEMIN
back-off interval is proportional to the difference be- to PAUSEMAX seconds and then randomly select another
tween the received signal power and an optimal value. destination and moving speed.
5) Upon receiving a duplicated RREQ packet, the node Trafﬁcs are generated as Constant Bit Rate (CBR) with 512
compares it with the previous one. If found it is better, bytes user data in each packet. Every trafﬁc link will transfer
it will update its routing table. 100 packets at a rate of 4pakcet/sec.
6) A node should drop all the RREQs that come after the The transmitter power level is rectiﬁed into 4 stages: 1 (level
back-off interval. 4), 0.5 (level 3), 0.2 (level 2), and 0.1 (level 1). We set the
We can assert that an optimal distance will result to shortest radio path loss factor α to be 3 and assume 200m can be
back-off interval; therefore this routing scheme will establish covered by the maximum power (level 4). Thus by the power
an optimal path with least duration. levels we set, their radio coverage radius are 200m, 158.74m,
119.96m, and 92.83m, respectively.
For each simulation scenario, we set the node three different
IV. ENERGY EFFICIENCY ANALYSIS AND initial energy variables for original AODV routing without
SIMULATION RESULTS hop-by-hop power control, original AODV routing with hop-
The power control routing scheme proposed at here cannot by-hop power control (noted as Original+ later in tables), and
increase the network capacity because maximum power is used power control routing for AODV we described in this paper,
in RTS/CTS handshake. This is necessary to avoid collisions. respectively. The simulation time is set to be 50sec.
Original Original+ PCR Average Hop Count Latency (ms)
20 nodes 0.0878 0.0528 0.0486 Original PCR Original PCR
40 nodes 0.0958 0.0605 0.0519 20 nodes 2.55 2.90 1.28 1.95
60 nodes 0.1074 0.0707 0.0589 40 nodes 3.73 3.95 1.86 3.84
80 nodes 0.1246 0.0861 0.0709 60 nodes 2.87 3.22 1.44 3.15
TABLE I 80 nodes 3.71 3.84 1.85 3.76
AVERAGE N ODE E NERGY D RAIN TABLE III
ROUTE E STABLISHMENT L ATENCY
Original Original+ PCR
20 nodes 0.0488 0.0323 0.0308 Original PCR
40 nodes 0.0923 0.0728 0.0561 20 nodes 35.0 29.1
60 nodes 0.0759 0.0530 0.0467 40 nodes 67.5 51.4
80 nodes 0.1007 0.0724 0.0535 60 nodes 128.9 95.7
TABLE II 80 nodes 194.9 130.3
S TANDARD D EVIATION OF E NERGY D RAIN TABLE IV
AVERAGE L INK B REAKS
In this model we don’t consider the energy consumptions
taken at MAC and network layer, such as RTS/CTS hand- delay. The average latency can be compared in Table. III.
shakes, broadcasting of RREQ and replying RREP, etc. They Because the power control routing protocol involves more
are not signiﬁcant comparing to the trafﬁc data size. nodes in routing, and back-off time introduced at intermediate
B. Simulation Results and Discussion nodes, the route establishment latency is about twice as that
of original AODV protocol.
Power Consumption and Standard Deviation: First of all,
Link Stability: When a node involved in a route has moved
we need to check the average energy drain after the 50-
out of the radio coverage of precursor node, a link break
second simulation. Table. I shows the result of average energy
occurs. Under this circumstance a new route establishment ap-
drain of each time. We can see that our scheme costs only
proach is triggered by the precursor node sending Route Error
approximately 55% of energy compare to the original AODV
RERR packet. Frequent link breaks degrade the transmission
routing in all the scenarios. One can check that it consumes
performance because they cause both transmission delay and
82-90% of energy if compare to that hop-by-hop power control
extra overheads. In our protocol, since it always selects the
is implemented only.
nodes with a distance around 100m as optimal candidates,
Another important feature is standard deviation of energy
the link break probability is signiﬁcant less than that of the
drain after the simulation. A lower standard deviation means
original AODV. Table. IV shows the simulation results of link
that each node has more equally involved in network activity
break counts of different scenarios. We have set a moderate
thus the overall network lifetime will be longer. Some liter-
mobility of network with randomly 1-10m/s moving speed.
atures such as  have been published for this topic and
assert that a mean energy drain among nodes will increase the
network lifetime. From Table. II we can see our scheme gives
much better standard deviation of energy drain than that of
the original AODV protocol. In this paper we have demonstrated our approach to energy
Route Establishment Latency: Since our protocol deploys conservation for ad hoc routing. In a reactive routing protocol,
back-off of RREQ relaying at intermediate nodes. It will it is possible to select a more energy efﬁcient route by
increase the route establishment latency. When a node has inspecting the hop-by-hop transmitting and received power
received a RREQ, it takes some time to process the packet when route request packet is broadcast throughout the network.
as well, so the delay on a node consists of two parts: back- With MAC layer power control, a signiﬁcant amount of en-
off time due to the power calculation and processing delay. ergy can be saved compare to that non-power-control scheme.
Together with propagation delay on radio link, the latency With power-aware routing protocol presented in this paper,
over one hop is the sum of these three factors. The processing we obtain even better energy conservation. In our protocol,
time and propagation of RREQ packet can be taken in some since a long hop is trend to be divided into shorter multihops,
hundreds microseconds. Here we set it as 500us. Therefore we more nodes are involved into the network activity evenly.
set our back-off unit to be 500us as well to ensure the routing This makes energy drain even in the network and increase
protocol. The back-off time is calculated by power level as: the overall network lifetime.
With proper distance of a hop, the link can endure more
Tbk = c level − 2 here c = 500µs (5)
node mobility and the probability of route broken is reduced.
By this formula, the latency at one node can be 1500us (when This will increase the transmission performance and reduce
power level is 4) together with processing and propagation the network overhead.
The power control routing scheme we proposed here is
simple to be implemented on any reactive ad hoc routing
protocol, even though we deployed AODV in our simulation.
The scheme also requires that transmitter power scalable. This
is not a technique problem because there are products available
for this purpose.
 Swetha Narayanaswamy, Vikas Kawadia, R.S. Screenivas and P.R. Kumar,
Power Control in Ad-hoc Networks: Theory, Architecture, Algorithm and
Implementation of COMPOW Protocol, In Proceedings of the European
Wireless Conference - Next Generation Wireless Networks: Technologies,
Protocols, Services and Applications, pages 156-162, Florence, Italy, Feb.
 P. Bergamo, D. Maniezzo, A. Giovanardi, G. Mazzini, M. Zorzi, Dis-
tributed Power Control for Power-Ware Energy-Efﬁcient Routing in Ad
Hoc Networks, EW2002, Florence, Italy, pp. 237-243.
 Eun-Sun Jung, Nitin H. Vaidya, A Power Control MAC Protocol for Ad
Hoc Networks, MOBICOM’02 September 23-28, 2002, Atlanta, Georgia.
 S. Doshi, S. Bhandare, T. X Brown, An On-demand minimum energy
routing protocol for a wireless ad hoc network. Mobile Computing and
Communications Review, vol. 6, no. 2, July 2002.
 Agarwal, S.; Katz, R.H.; Krishnamurthy, S.V.; Dao, S.K., Distributed
power control in ad-hoc wireless networks. Personal, Indoor and Mobile
Radio Communications, 2001 12th IEEE International Symposium on ,
Volume: 2 , 30 Sept.-3 Oct. 2001.
 S.-J. Park and R. Sivakumar, Load Sensitive Transmission Power Control
in Wireless Ad-hoc Networks, IEEE Global Communications Conference
(GLOBECOM), Taipei, Taiwan, November 2002.
 C. E. Perkins and E. M. Royer. Ad hoc On Demand Distance Vec-
tor (AODV) Routing. IETF Internet Draft, draft-ietf-manet-aodv-13.txt,
 D. JOHNSON, D. MALTZ, Y-C. HU AND J. JETCHEVA, The Dynamic
Source Routing (DSR) Protocol for Mobile Ad Hoc Networks, Internet
Draft, draft-ietf-manet-dsr-05.txt, work in progress, June 2001.
 IEEE 802 LAN/MAN Standards Committee, Wireless LAN medium
access control (MAC) and physical layer (PHY) speciﬁcations, IEEE
Standard 802.11, 1999 edition, 1999.
 Cano, J.-C.; Manzoni, P., A performance comparison of energy consump-
tion for Mobile Ad Hoc Network routing protocols, Modelling, Analysis
and Simulation of Computer and Telecommunication Systems, 2000.
Proceedings. 8th International Symposium on, 29 Aug.-1 Sept. 2000,
Page(s): 57 -64
 Hong Jiang; Garcia-Luna-Aceves, J.J.; Performance comparison of three
routing protocols for ad hoc networks. Computer Communications and
Networks, 2001. Proceedings. Tenth International Conference on, 15-17
Oct. 2001, Page(s): 547 -554
 Woncheol Cho; Seong-Lyun Kim; A fully distributed routing algorithm
for maximizing lifetime of a wireless ad hoc network. Mobile and Wireless
Communications Network, 2002. 4th International Workshop on, 9-11
Sept. 2002, Page(s): 670 -674