. 58
( 87 .)


Scalable Routing Framework for Ad Hoc Networks,” in M. Ilyas (Ed.), Handbook of Ad Hoc
Wireless Networks, Chapter 14. CRC Press, 2002.
55. K. Sanzgiri, B. Dahill, B. N. Levine, C. Shields, and E. M. Belding-Royer. “A Secure Routing
Protocol for Ad Hoc Networks,” in Proceedings of the 10th IEEE International Conference on
Network Protocols (ICNP), Paris, November 2002.
56. N. Shacham. “Hierarchical Routing in Large, Dynamic Ground Radio Networks,” in Proceed-
ings of the 18th Hawaii InternationalConference on System Sciences, pp. 292“301, 1985.
57. S. Singh, M. Woo, and C. S. Raghavendra. “Power-Aware Routing in Mobile Ad hoc Net-
works,” in Proceedings of the 4th Annual ACM/IEEE International Conference on Mobile Com-
puting and Networking (MobiCom), pp. 181“190, Dallas, Texas, October 1998.
58. I. Stojmenovic. “Location Updates for Efficient Routing in Ad Hoc Wireless Networks,” in
Handbook of Wireless Networks and Mobile Computing, pp. 451“471. Wiley, 2002.
59. I. Stojmenovic. “Position Based Routing in Ad hoc Mobile Networks.” IEEE Communications
Magazine, 40(7), 128“134, July 2002.
60. P. F. Tsuchiya. “The Landmark Hierarchy: A New Hierarchy for Routing in Very Large Net-
works,” in Computer Communication Review, pp. 35“42, Stanford, CA, August 1988.
61. Y. Xu, J. Heidemann, and D. Estrin. “Geography-informed Energy Conservation for Ad Hoc
Routing,” in Proceedings of the 7th Annual ACM/IEEE International Conference on Mobile
Computing and Networking (MobiCom), pp. 70“84, Rome, July 2001.
62. S. Yi, P. Naldurg, and R. Kravets. “A Security Aware Routing Protocol forWireless Ad Hoc Net-
works,” in Proceedings of the 6th World Multi-Conference on Systemics, Cybernetics and Infor-
matics (SCI), pp. 286“292, Orlando, FL, July 2002.
63. M. G. Zapata and N. Asokan. “Securing Ad Hoc Routing Protocols,” in Proceedings of the ACM
Workshop onWireless Security (WiSe), Atlanta, GA, September 2002.
64. J. Zavgren. “NTDR Mobility Management Protocols and Procedures,” in Proceedings of the
IEEE Military Communications Conference (MILCOM), pp. 292“301, November 1997.
65. Y. Zhang and W. Lee. “Intrusion Detection in Wireless Ad hoc Networks,” in Proceedings of the
6th Annual ACM/IEEE International Conference on Mobile Computing and Networking (Mobi-
Com), pp. 26“33, Seattle, August 1999.




One reason why working on energy-efficient communication in ad hoc networks is so
much fun is the complexity of trade-offs available to the designer of energy-aware sys-
tems. The richness of interactions among the physical elements of the system, the various
layers of the protocol stack, and the environment in which the system operates requires
creative and careful attention to obtain interesting and meaningful results.
This chapter surveys current work on energy-efficient communication in ad hoc wire-
less networks, focusing on problems and approaches that are most specific to the decen-
tralized ad hoc environment and illustrate most clearly its unique challenges. In addition, I
have chosen to emphasize practical issues and approaches and to focus on work that is
largely based on readily available hardware and communication technology.
The chapter opens with a brief introduction to ad hoc wireless networks and discusses
characteristics that make these networks structurally different from infrastructure wireless
networks, necessitating the development of new energy management techniques.
The second section motivates the goal of energy-efficient communication by present-
ing some results obtained by measuring the energy consumption of various (mostly IEEE
802.11-based) devices. One key result is that the energy consumed by an idle network in-
terface dominates total energy consumption.
The third section examines some existing and proposed power save protocols. Power
save protocols attack the problem of high energy consumption in an idle network interface
by selecting intervals during which the interface can use a low energy consumption sleep
state, with minimal impact on overall network performance.

Mobile Ad Hoc Networking. Edited by Basagni, Conti, Giordano, and Stojmenovic.
ISBN 0-471-37313-3 © 2004 Institute of Electrical and Electronics Engineers, Inc.

The fourth section presents the topology control and minimum energy routing prob-
lems. In a multihop network, nodes can alter their transmit power level to manipulate the
effective network topology, reducing interference and increasing network capacity, as well
as reducing energy consumption.
The fifth section discusses maximum lifetime routing. In an ad hoc network, nodes co-
operate to forward traffic from a source to a destination. A node that forwards traffic on
behalf of other nodes may exhaust its energy reserves, so that it can no longer participate
in the network. It is therefore necessary to select paths in a way that maximizes the net-
work lifetime.
The appropriate metric for network lifetime depends to some extent on the application
scenario. In a sensor network, data is forwarded to distinguished gateway nodes, which
are the only destinations in the network. Such a network is usually modeled as a dense,
uniform collection of functionally equivalent nodes: As long as sensing and communica-
tion coverage is maintained, the lifetime of any individual sensor is relatively unimpor-
tant. In a personal communication network, nodes are devices associated with individuals.
This means that any node can be a destination and that loss of connectivity to any node is
There are two recurring themes in this chapter. The first is that problems of energy ef-
ficiency cannot be isolated to a single layer in the protocol stack. The second is the extent
to which wireless propagation and energy consumption models affect the design and eval-
uation of energy-efficient techniques. Examples appear throughout the chapter, which
closes with some discussion of these themes.


In an ad hoc wireless network, nodes cooperatively form a network independently of any
fixed base station infrastructure. These networks are generally characterized by band-
width-constrained, variable-capacity links and an unpredictable, dynamic topology. Each
node communicates directly with destinations within wireless transmission range and in-
directly with all other destinations, relying on its peers to forward traffic on its behalf. Ad
hoc routing is an active area of research [31] and we now find a complete alphabet soup of
proposed routing protocols, ranging from AODV [28] to ZRP [15]. The nodes of an ad
hoc network also cooperate to provide application services such as service discovery,
namespace and session management, and security [14].
Because the nodes of an ad hoc network are usually small, battery powered devices,
energy management is a critical issue for practical deployment of these networks.1 Ad hoc
networks differ from wireless infrastructure networks in two fundamental ways that re-
quire unique strategies to obtain energy-efficient behavior.
First, infrastructure wireless networks have a strongly asymmetric structure. Although
the mobile units are battery constrained, the base stations have no such limitation. Low-
level energy management strategies are often based on spending energy at the base station
to conserve energy at the mobile node. This is not a viable approach in an ad hoc network,
which has no such fixed elements.
Second, in an infrastructure wireless network, nodes operate independently of each
other, using the base station to communicate with other nodes and access services in the

Vehicle-based ad hoc networks are a significant exception.

infrastructure network. Energy management strategies therefore only need to consider the
node and its local applications. In an ad hoc network, nodes are highly interdependent and
must cooperate to provide routing and other services, making it important to maximize
the network lifetime. Greedy strategies, in which each node seeks only to minimize its
own energy consumption, are not effective in this context.


Before considering various approaches to reducing communication energy consumption,
it is reasonable to question whether the network interface contributes significantly to the
overall energy consumption of a mobile system.
The variety of devices, operating modes, energy management techniques, and usage
scenarios make it impossible to make blanket statements about energy consumption in
portable devices. Obviously, measurements of specific systems quickly become outdated.
Nevertheless, measurements [35] show that the network interface represents a significant
fraction of the energy consumed by a laptop PC and is dominant source of energy con-
sumption in some PDA hardware. More recently, in a preliminary implementation of a
Bluetooth-based sensor device [12, 18], the interface accounted for over 40% of the total
energy consumption when the Bluetooth device was in standby mode. Moreover, the rela-
tive cost of communication may be expected to increase as advances continue to be made
in low-power hardware and energy-efficient operating systems and applications. This
trend will accelerate as communication functionality is increasingly incorporated into
small, specialized devices such as sensors.
The design and evaluation of energy-efficient communication protocols therefore re-
quires practical understanding of the energy consumption behavior of the underlying net-
work interface. The energy consumed by an interface depends on its operating mode: In
the sleep state, an interface can neither transmit nor receive, so it consumes very little en-
ergy. To be able to transmit or receive, an interface must explicitly transition to the idle
state, which requires both time and energy. In the idle state, an interface can transmit or
receive data at any time, but it consumes more energy than it does in the sleep state, due to
the number of circuit elements that must be powered.
Because of their wide availability, low-cost and relatively stable, open specification,
IEEE 802.11-based [16] interfaces have attracted considerable attention. Table 11.1 sum-
marizes some experimental measurements of the power consumption of various network
interfaces. Although the data vary somewhat among the various manufacturers, models,
and measurement methods, there are consistent patterns. Transmitting requires more ener-

Table 11.1. Some Power Consumption Measurements
Interface Transmit Receive Idle Sleep Mbps
IEEE 802.11 Interfaces (2.4 GHz)
Aironet PC4800 [8] 1.4“1.9 W 1.3“1.4 W 1.34 W 0.075 W 11
Lucent Bronze [10] 1.3 W 0.97 W 0.84 W 0.066 W 2
Lucent Silver [10] 1.3 W 0.90 W 0.74 W 0.048 W 11
Cabletron Roamabout [6] 1.4 W 1.0 W 0.83 W 0.13 W 2
Lucent WaveLAN [20] 3.10 W 1.52 W 1.5 W ” ”

send send
1500 RTS data

recv recv

50 mV/div = 50 mA/div
1000 idle idle

power (mW)




500 usec/div

Figure 11.1. Unicast transmission”256 bytes at 2 Mbps (Lucent IEEE 802.11 card).

gy than receiving, but the difference is much less than a factor of two. The idle energy
consumption is quite high, comparable to that of receiving and an order of magnitude
more than that of sleeping.
More detailed results are presented in [10, 8], which describe direct measurements of
a network interface card as it transmits and receives packets of varying sizes, as in
Figure 11.1. In [10], this data is used to develop a packet-level energy consumption
model for a Lucent IEEE 802.11 network interface. The energy consumed (in addition
to the baseline idle energy consumption) when the interface transmits, receives, or dis-
cards a packet can be described using, at each operation, an incremental component that
is proportional to the size of the packet and a fixed component that reflects channel ac-
quisition and other overhead. (The effects of contention and retransmissions are not ad-
dressed in this work.)
Although the specific numerical results are of limited value, again there are consistent
patterns. The fixed overhead is high, due to the cost of the RTS/CTS/data/ACK handshake
and the size of the IEEE 802.11 MAC headers. The fixed overhead dominates for packets
smaller than 338 bytes (2 Mbps) or 1.2 Kbytes (11 Mbps).2 Nevertheless, it is the energy
consumption of the idle state that dominates total energy consumption. A rough calcula-
tion based on [10] shows that an IEEE 802.11 interface sending 10 128-byte broadcasts (2
Mbps) per second and receiving the same from each of four neighbors consumes only
about 1% more power than an idle interface.
The results obtained from such experiments can be incorporated into packet-level sim-
ulations of ad hoc routing protocols, as described in [9]. It is clear that bandwidth and en-
ergy are not analogous metrics: minimizing bandwidth usage does not necessarily mini-
mize energy consumption. In particular, the results show that in a moderately dense
network, broadcast traffic can be expensive, due to the multiplied cost of receiving.
Promiscuous mode operation is similarly costly.
The experiments described in [10] did not measure the effects of transmit-power con-
trol on the energy consumption at the network interface. Measurements of the Aironet
PC4800B card, which supports multiple transmit-power levels, are reported in [8]. The re-

The difference is due to the fact that IEEE 802.11 control traffic is transmitted at the (slower) base data rate.

sults show that as the the output power level decreases from 50 mW to 1 mW, the power
consumed by the transmitter decreases about 500 mW. This variation represents only
about 25% of the total power consumption; the baseline power consumption accounts for
about 70% of the total. The power consumption of the receiver is comparable to the base-
line and roughly independent of the transmit-power level.
The conclusion to be drawn from these data is obvious. To reduce the energy consump-
tion of the network interface, it is necessary to find a way for the interface to spend more
time in the sleep state and less time awake in the idle state. Such power-save protocols are
the topic of the next section.


A power-save protocol puts a node™s network interface into the sleep3 state in order to save
energy. A sleeping node cannot forward or receive traffic and its unavailability may inter-
rupt the flow of traffic though a multihop ad hoc network. Power-save protocols therefore
seek to maximize energy saving, while minimizing impact on throughput, latency, and
route latency. Two main classes of power-save protocols, network layer protocols and
MAC layer protocols, are introduced below.

11.4.1 Network-Layer Power-Save Protocols
Network-layer protocols make up the largest class of power-save protocols. Scheduling of
the interface is driven by network-layer traffic, which is buffered at the MAC layer for
sleeping neighbors or routed so as to take advantage of nonsleeping ones. Power-save pro-
tocols are based on three basic strategies, outlined below:


. 58
( 87 .)