Computer Networks 50 (2006) 29022917

www.elsevier.com/locate/comnet

Geographic routing in the presence of location errors

Sungoh Kwon, Ness B. Shro

Center for Wireless Systems and Applications (CWSA), School of Electrical and Computer Engineering, Purdue University,
1285 EE Building, West Lafayette, IN 47907-1285, United States
Received 2 June 2005; received in revised form 10 October 2005; accepted 17 November 2005
Available online 28 December 2005
Responsible Editor: E. Ekici

Abstract
In this paper, we propose a new geographic routing algorithm that alleviates the eect of location errors on routing in
wireless ad hoc networks. In most previous work, geographic routing has been studied assuming perfect location information. However, in practice there could be signicant errors in obtaining location estimates, even when nodes use GPS.
Hence, existing geographic routing schemes will need to be appropriately modied. We investigate how such location
errors aect the performance of geographic routing strategies. We incorporate location errors into our objective function
by considering both transmission failures and backward progress. Each node then forwards packets to the node that maximizes this objective function. We call this strategy Maximum Expectation within transmission Range (MER). Simulation results with MER show that accounting for location errors signicantly improves the performance of geographic
routing. Our analysis also shows that our algorithm works well up to a critical threshold of error. We also show that
MER is robust to the location error model and model parameters. Further, via simulations, we show that in a mobile environment MER performs better than existing approaches.
 2005 Elsevier B.V. All rights reserved.
Keywords: Wireless ad hoc networks; Geographic routing; Location errors; Simulations

1. Introduction
Geographic routing for multi-hop wireless networks has become an active area of study over the
last few years (e.g., see [14] and the references
therein). Geographic routing is appealing because
q

This work has been partially supported by the NSF grant

ANI-0207728 and a grant from the Indiana 21st Century Fund.
*
Corresponding author. Tel.: +1 765 494 3471; fax: +1 765 494
3358.
E-mail addresses: [email protected] (S. Kwon), shro@
purdue.edu (N.B. Shro).

of its simplicity and scalability. However, most

work in this area has implicitly assumed that location information available at each node is perfect,
while in practice only a rough estimate of this information is available. In this paper, we will show that
imperfect location information can lead to substantial degradation in the performance of geographic
routing. We will also develop a routing scheme that
accounts for location errors, and whose performance is robust to such errors.
In geographic routing, each node determines its
own location by using either the Global Positioning

1389-1286/$ - see front matter  2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.comnet.2005.11.008

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

System (GPS) [5,6] or the location sensing techniques [710]. It then broadcasts its location information to other nodes proactively and periodically.
Packet forwarding is accomplished based on the
neighbors location information stored in each
nodes database (DB) and the destinations location
information contained in the packet. Packets are
typically forwarded using what is commonly referred
to as the greedy mode, in which nodes use local information to forward packets towards their destination.
If the greedy mode is not successful (i.e., the destination node is not available in the local databases, or a
greedy mode forwarding results in a failure), a special routine called a recovery mode is initiated
through the entire network to nd an appropriate
route to the destination. Since the greedy mode uses
local information and most packets are forwarded
in this mode [11], geographic routing is generally
considered to be scalable and applicable to large
networks.
Several forwarding schemes have been proposed
for use in the greedy mode [1215]. Fig. 1 provides
an illustration of such schemes when a node S with
transmission range R has a packet to send to some
_
_
node D. Arcs ab and pq are centered at node D
and with radii DS and DG, respectively. When the
nodes in the wireless network have a xed transmission range, the Most Forward within Radius
(MFR) scheme [12] and the Greedy Routing
Scheme (GRS) [13] have been proposed to minimize
the hop count and the energy consumption. MFR
forwards a packet to the neighbor (node M in
Fig. 1) that is the farthest from the source in the

a
p

M
C

G
D

Fig. 1. The examples of forwarding schemes when node S with

transmission range R has a packet to send to node D: compass
routing(C), GRS(G), MFR(M) and NFP(N).

2903

direction of the destination within the transmission

range. GRS selects the closest neighbor (node G in
Fig. 1) to the destination among neighbors. Since
in most cases MFR and GRS provide the same path
to the destination [16], we only consider GRS in this
paper. When nodes have the ability to control the
transmission ranges, the Nearest Forward Progress
(NFP) algorithm [14] has been proposed to reduce
energy power consumption. NFP chooses the closest neighbor (node N in Fig. 1) to the sender within
the forward region. Yet another scheme called compass routing [15] is used to select that neighbor
(node C in Fig. 1), which has a minimum angle with
respect to the line between the sender and the
destination.
Most work on geographic routing assumes that
location information received from GPS (or other
techniques) is perfect [1,1620]. Eective algorithms
in the recovery mode have been addressed in [1,16
19]. Flooding based algorithms have been proposed
for nding alternative paths in [16,19]. Face routing
(perimeter routing) has also been studied for route
recovery in [1,17]. In [20], the authors consider the
information discrepancy due to mobility and propose a mobility prediction scheme based on
reported location information. However, the location information in [20] is assumed to be measured
perfectly by GPS.
In practice, the received location information is
not perfect [5,10,21] and the error in the location
information degrades the performance of geographic routing [22]. This inaccuracy is caused even
when nodes use GPS because of GPSs inherent
error in location estimation [6]. This location error
could induce transmission failures and backward
progress in the greedy mode. A transmission failure
happens when the selected node is out of the transmission range. Backward progress occurs when the
chosen node is located farther from the destination
than the sender and could cause looping. Loops
occur when the selected node is one of the previous
senders in the route. Further, it could also result in
a local minimum, i.e., there exist solutions but
there are no nodes in the forwarding direction.
Such failures produce unnecessary transitions from
the greedy mode to the recovery mode, which in
turn results in an inecient routing solution
[16,17].
In this paper, we study the impact of these location errors on the performance of geographic routing. We further propose a new routing scheme to
improve the performance of geographic routing.

2904

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

We focus on the case when the transmission ranges

of the nodes are xed. Using numerical simulations
we verify the performance of the proposed algorithm and the robustness of the location error models and model parameters. Unless otherwise stated,
the term location error means location error
due to measurement through this paper.
The rest of the paper is organized as follows. In
Section 2, we study how location errors in geographic routing arise and evaluate the impact of
location errors on the performance of the geographic routing. In Section 3, we propose a new
algorithm for geographic routing in an environment
with location errors and analyze properties of the
algorithm. In Section 4, we use simulation to compare the performance of our scheme with that of
known schemes. Section 5 concludes this paper.
2. The impact of location error on geographic
routing performance
In this section, we develop a location error
model and investigate the impact of location errors
on the performance of geographic routing. As
mentioned in the introduction, errors in the location information aect the forwarding scheme in
the greedy mode, which could cause unnecessary
transitions into the recovery mode of the algorithm
in order to nd an alternative route to the destination. The goal of the greedy mode is to succeed in
transmitting packets to a neighbor with forward
progress (i.e., the neighbor is closer to the destination). Thus, it is important to account for both
transmission failures and backward progress to analyze how the location error aects the performance
of geographic routing.
2.1. Error modeling
Location errors occur during the process of estimating the location (via GPS or other techniques)
[5,6,10,21]. In GPS, the performance of an estimated
location depends on the geometry of satellites, location sensing techniques, radio environments, etc.
[5,23,24]. For example, when GPS uses a single frequency, the root mean square measurement error is
typically 6 m [6]. On the other hand, the typical measurement error is about 3 m for a dual-frequency
receiver [6]. The geometry of satellites and location
sensing techniques are known parameters to the
GPS receiver. Other factors are also estimated and
adjusted by the GPS receiver.

Xj

z
Xi
Xi

Xj

Destination
Node d
Xd

ri

measured location
real location

Fig. 2. Location error modeling when node i with the transmission range ri has a packet to send the destination node d.

In this paper we make the following assumptions.

All nodes are equipped with GPS1 to measure their
own locations. These locations are proactively
broadcasted. The location errors at dierent nodes
are independent. The location error at each node
is modeled by a Gaussian distribution2 with zero
mean and nite standard deviation. The zero-mean
assumption implies that, for an given environment,
the average of location errors over all nodes is equal
to zero, i.e., it follows fromPthe strong law of large
n
numbers [25] that limn!1 1n k1 W k is equal to zero
almost surely, where Wi represents a measurement
error of node i. However, note that for a given environment, the time average of location errors at a single node could be non-zero. Therefore, a scheme that
uses simple averaging of samples in time at each
node will not be able to overcome the errors.
Let Xi be the real position of node i and let X 0i be
its measured position. Then Xi can be expressed as
X i X 0i W i , where Wi is a Gaussian random
vector with zero mean and standard deviation ri.
For convenience, we assign node i to be at the
origin and destination node d to be on the x-axis,
as in Fig. 2. Let node j be a neighbor of node i,
and let z be the real distance between the two nodes.
Since Xi and Xj are independent and Gaussian, the
probability density function fj(z, h) that node j is
located at (z, h) is
1
Our analysis and algorithm are applicable to other location
estimation technologies, however in this paper we focus on
networks where the nodes are equipped with GPS.
2
In Section 4, we will also study cases when the errors are not
Gaussian and their impact on the routing performance.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

!
!
z2 g2ij
zgij
z
exp
fj z; h
exp 
cos h ;
2pr2ij
2r2ij
r2ij
where rij is the standard deviation of Xi  Xj, h is an
angle of Xj with respect to x-axis, and gij kX 0i 
X 0j k as in Fig. 2. Hence the probability density
function f(z) that the distance between two nodes
is z is
Z 2p
fj z
fj z; hdh
0
!
!
z2 g2ij
zgij
z
2 exp 
I0
for z P 0;
rij
2r2ij
r2ij
where I0(x) is the modied Bessel function of the
rst kind R and zero order, which is dened by
2p
1
I 0 x 2p
expx cos h dh.
0
2.2. Transmission failure probability
A packet transmission failure occurs when a chosen node is out of the transmission range of a sender, as shown in Fig. 3. Most work assumes that
the transmission range of each node is perfectly circular and identical so that a neighbor is within the
transmission range of a node if and only if the node
is located within the transmission range of the
neighbor. In practice, nodes have imperfect circular
transmission patterns [26] and the transmission
ranges deviate from the ideal case [20]. Moreover,
a network may be composed of heterogeneous

2905

nodes that have dierent transmission ranges. In

these cases, even though a node is located in the
transmission range of a neighbor, the neighbor
may be out of the transmission range of the node.
Such a link is an asymmetric communication link.
In the case of asymmetric communication links,
transmission failures can happen in the presence of
location errors even though each node precisely
knows its own transmission range and pattern. In
a mobile environment, displacements of nodes
induced by mobility cause transmission failures that
displacement prediction may reduce [20,27]. However, this is still vulnerable to the transmission
failure due to location errors in measurement even
though the prediction is perfect and the communication links are symmetric. In the case when the
transmission range is controllable, the adjusted
transmission range of a sender can also fail to transmit a packet to its neighbor that is still within the
maximum transmission range of the sender. Here,
we study failures caused only by inaccurate location
information.
Assume that node i has a packet to transmit and
node j is chosen as the next node. We assume that
node i calculates the distance between nodes i and
j as gij such that gij kX 0i  X 0j k, and sets its transmission range to ri, which may be xed or controllable, in order to forward data to node j. The
probability that a packet transmission from node i
to node j fails is
Prftransmission failure at node jg PrfZ > ri g
Z 1
f zdz

ri

!
!
z2 g2ij
zgij
z

exp 
I0
dz
r2ij
2r2ij
r2ij
ri
!! 

Z 1
gij
1 2 g2ij
r 2
r exp 

dr
I0 r
ri
2
rij
rij
rij


gij ri
;
Q1
;
rij rij
Z

Heterogeneity

Irregular Tx pattern
Node displacement
Tx failure

Controlled Tx range

Displacement

Ideal (or max.) Tx range

Real (or controlled) Tx range

Real location
Measured location

Fig. 3. Examples for transmission failure.

where Z = kXi  Xjk, rij is the standard deviation of

Xi  Xj, and Q1(a, b) is a Marcums Q function with
m = 1 dened as in [28].
It follows from Marcums Q function with m = 1
that the transmission failure probability increases
when the standard deviation of location errors rij
increases. When rij is xed and the chosen node is
closer to the edge of the transmission range, the
transmission failure probability increases. In other

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

words, given an error environment, a longer transmission range reduces the transmission failure probability.
2.3. Backward progress probability
Backward progress occurs when a chosen node j
is located farther from a destination than the sending node i. Note that there may exist a route to the
destination even though there is no neighbor in the
forward region. This case is typically called a local
minimum in the geographic routing literature and
cannot be solved by using only the greedy mode,
so recovery mode is needed. Assume that node i
has a packet to transmit and node j is chosen as
the next node. For simplicity, in this subsection,
we assume that the destination location Xd in the
packet has no error.3
The probability that the chosen node j, such that
kX 0i  X d k P kX 0j  X d k, is located behind (is
further away from the destination than node i) the
sender i is
Prfbackward progress at node jg

ation of location errors rij increases. When rij is xed

and the chosen node is closer to the sender, the
backward progress probability increases.
Figs. 4 and 5 show these two probabilities. In
Fig. 4, we x the location of node j and show the
relationship between the failure probabilities and
the standard deviation of the location error. In
Fig. 5, we x the location error and show the relationship between the failure probabilities and the
distance between node i and node j. Given the location error, when the chosen node is closer to the

1
Transmission failure probability
Backward progress probability

0.9
0.8
0.7
Failure probability

2906

0.6
0.5
0.4
0.3

Prfz; hjkX i  X d k 6 kX j  X d kg
Z

fj z; h dz dh

0.2
0.1

fz;hjkX i X d k6kX j X d kg

fz;hjkX i X d k6kX j X d kg

z2 g2ij
 exp 
2r2ij

z
2pr2ij

!
zgij
cos h dz dh;
exp
r2ij

where gij kX 0i  X 0j k and rij is the standard deviation of Xi  Xj.

In general, the integral above does not reduce to
a closed form and must be numerically evaluated.
However, if ri is much smaller than the distance
between node i and the destination, h is close to 0,
we can approximate Eq. (2) as follows:
 
gij
Prfbackward progress at node jg Q
;
rij
R
1 exp 
where gij kX 0i  X 0j k and Qx p1

2p x
2
 x2 dx.
It follows from the Q function that the backward
progress probability increases when the standard deviThe case where the destination location is in error can be
similarly treated, although the equations become more notationally complex. For the simulations, we assume that the location
information of all nodes, including the destination, has errors.

0.2
0.4
0.6
0.8
Normalized standard deviation of location errors

1
Transmission failure probability
Backward progress probability

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Fig. 4. Failure probabilities versus the standard deviation of

location errors when gij = 0.8ri. The standard deviation of
location errors is normalized by the transmission range ri.

Failure probability

0.2

0.4
0.6
Normalized location from a sender

0.8

Fig. 5. Failure probabilities versus the location of node j when

the standard deviation of location errors is 0.1ri. The location of
the chosen node is normalized by the transmission range ri.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

sender or the edge of the transmission range, the

failure probability increases.
2.4. Impact of location error on geographic routing
All forwarding schemes in geographic routing
suer from the above mentioned failures. GRS (or
MFR) selects the closest neighbor to a destination
(or the farthest neighbor from the source in the
direction of the destination), so the node is more
likely close to an edge of a transmission range than
any other neighbors. They are susceptible to a transmission failure. NFP chooses a neighbor which is
closest to the sender. This scheme is susceptible to
backward progress. Compass routing does not consider the distance between the sender and the intermediate node, but cares for only the angle with
respect to the line between the sender and the destination. Hence, compass routing is vulnerable to
both factors described above. When the number of
nodes increases in a given area, a chosen node is closer to the sender or the edge of the transmission
range. Therefore, denser nodes can potentially
worsen the performance (as will also be shown via
numerical studies in Section 4).
3. Geographic routing scheme with location errors
In this section we propose a new geographic routing scheme that can mitigate the impact of location
errors. Since MFR is similar to GRS in most cases
[4], we focus on improving GRS. For ease of illustration, from here on we assume that the transmission range is xed. However, it should be readily
apparent that the methodology can be extended to
the case when the transmission range is controllable.
Since each node measures its location and estimates
its own error characteristic, we attach an error
information eld in a message for geographic routing and announce the statistical characteristics of
the location error to neighbors with location
information.
3.1. Objective function
Let nodes i and j be located at Xi and Xj, respectively. We assume that X 0i and X 0j are the measured
locations of nodes i and j, respectively. As before,
these are expressed as X i X 0i W i and X j
X 0j W j , where Wi and Wj are Gaussian random
vectors with zero means and standard deviations
ri and rj, respectively. Then the real position of

2907

node j with respect to node i is a Gaussian random

vectorqwith
mean X 0j  X 0i and standard deviation

rij r2i r2j . Hence, the probability that node j

is located within uj from X 0j  X 0i is PrfkX 


u2j
0
0
X i  X j k 6 uj g 1  exp  2r2 , where X is the
ij

real position of node j with respect to node i.

Fix a sender i and the destination d. We dene
measured progress to node j to be tj kX d k
kX d  X 0j  X 0i k and the measured margin from
the boundary to be sj ri  kX 0i  X 0j k, where ri is
the transmission range of node i and Xd is the destination position with respect to node i. Then we can
express GRS in this (location) error-free environment as follows. If kX 0i  X d k 6 ri , choose node d.
If kX 0i  X d k > ri , choose node k such that
3

k arg max tj
j2N i

subject to kX 0i  X 0j k 6 ri

8j 2 N i ;

where Ni is the set of neighbors of node i.

When we have location errors, instead of using
the measured progress tj in (3), we propose using a
dierent metric to determine which neighbor to
forward the packets.
We dene true progress to node j to be sj
kX d k  kX d  X j  X 0i k when node j is actually
located at position Xj. Then, sj(Xj) is a random
variable with probability density function fj (Xj).
Since the probability density function, fj (Xj),
that node j is located at Xj is circularly symmetric
with respect to the point X 0j  X 0i , we consider area
Aj such that Aj fX 2 R2 jkX  X 0j  X 0i k 6 uj g
and the expected progress of node j over Aj as
follows.
Z
sj X j fj X j dX .
4
Efsj X j 1Aj g
Aj

If uj = 1, (4) becomes the expected progress over

the entire domain of X. However, if uj > sj, node j
may be out of the transmission range of node i. If
uj > tj, node j may be located behind node i. In order
to nd a neighbor to be able to successfully transmit
to and result in forward progress, we let uj =
min{sj, tj}.
For simplicity, we use an approximation of (4).
For realistically sized wireless networks, in most
_
cases kXdk is much greater than ri. Then, arc AB
_

and arc MN in Fig. 6 are nearly straight. Since

fj(z) is circularly symmetric, we can simplify (4) as
follows:

2908

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

Xj

M
sj
u j X'X'
j
i

ri

N
Node

Destination
D
Xd

O
tj

3.2. Properties of the algorithm

Fig. 6. The expected progress of node j with respect to node i that

has the transmission range ri and is located at O.

Efsj X j 1Aj g tj

follows: if the destination node exists in the neighbor list, the sender forwards the packet to the destination without any eort. However, since in the
location error environment the destination may be
located out of the transmission range, it is imperative to consider the parameter di to increase the
transmission success rate. In Fig. 7, we summarize
our algorithm. For the simulations in Section 4,
we also add functionalities to the protocol to detect
loops and local minima.

fj X j dX j ;
Aj

where Aj fX 2 R2 jkX  X 0j  X 0i k 6 uj g for u =

min{sj, tj}.
Hence, we dene Ej as the revenue of node j as
follows.
Ej ,tj F j uj ;

where
F j uj PrfkX  X 0j  X 0i k 6 uj g
!
u2j
1  exp  2
2rij
for uj = min{sj, tj}.
Based on the calculated revenue of each node,
node i selects the next node which has a Maximum
Expectation within transmission Range ri (MER).
Our MER algorithm for forwarding packets is given
as follows. If kX 0i  X d k 6 ri  di , where 0 6 di < ri,
choose node d. If kX 0i  X d k > ri  di , choose node
k such that
k arg max Ej
j2N

subject to kX 0i  X 0j k 6 ri

8j 2 N i ;

where Ni is the set of node is neighbor nodes stored

in the node is DB. Here, di is a function of the location errors and could be a tuning parameter for specic implementations. di can be simply the standard
deviation of location errors or the distance from the
maximizer of the objective function to the transmission range edge. In [20], the authors proposed a
scheme to improve the forwarding performance as

We now analyze the MER algorithm. For simplicity, we assume that a selected node is located
on the line between the sender and the destination
since the node on the line (OD in Fig. 6) is the
most likely chosen among neighbors that have the
same measured progress. Let the transmission range
of the sender ri be 1. We let r denote the standard
deviation for the overall location error between
the sender and the intermediate node. Then, (5)
becomes



u2
Et tF u t 1  exp  2
;
6
2r
where t 2 (0, 1) is the position of the selected node
and u = min{t, 1  t}.
We dene an eective search range to be an area
from the source to the most likely position chosen
by the algorithm and tmax to be a maximizer of
E(t) such that E(tmax) P E(t) for t 2 (0, 1).
Property 1. Let r  1, then MER is identical to
GRS within an effective search range for a next node.
The range is reduced by d such that E 0 (1  d) = 0,
where E 0 (t) is the derivative of E(t) and the maximizer
of E(t), tmax, is a decreasing function of r.
Proof.
the exponential term in (6),
 Since

u2
exp  2r
,
goes
to zero faster than r2 when r goes
2
to zero, for r  1, we have
Et t or2 .

Note that (7) is equivalent to the objective function

of GRS (3). Since E 0 (t) = tF(t) is always an increasing function of t 2 (0, 0.5), we need to consider
t 2 (0.5, 1) to check the maximum of E(t). The derivative of the objective function E 0 (t) for t 2 (0.5, 1) is
1
E0 t F 1  t  tf 1  t < 1  f 1  t.
2

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

2909

A packet arrives at node i

No

||Xi'-Xd|| > ri-i

Yes
Calculate tj = ||Xd-Xi||-||Xd-Xj||
for all neighboring nodes

Choose node d

Calculate Ej = tj Fj
for all neighoring nodes

Choose k = arg max Ej

Yes

Succeed
in transmiting?

Succeed in delivering the

packet to the destination

No

Succeed
in transmiting?

Fail to deliver the

packet to the destination

Yes

Set i = k.

Fig. 7. Block diagram for MER.

Since r  1 from the

above and f(t) has
 assumption

a maximum r1 exp  12 at t = r, E 0 (t)  1. The
values of E(t) precipitates from a maximum value
to zero, and the effective search range reduces to
tmax such that E 0 (tmax) = 0. In other words, the
effective search range is reduced by d such that
d = 1  tmax. Since r2 > 0 and
0

E t F 1  t  tf 1  t
F 1  t 1  tf 1  t  f 1  t
2

1  tmax
1  tmax
1  tmax


2
4
2r
8r
r2
4
3
1  tmax
1  tmax


1

t


max
2r4
2r2
!
4
1  tmax
o
;
4
r

we have
q
2
1  tmax
4  tmax 101  tmax
r2

4


1  tmax 4  tmax
o1  tmax 3 ;
4

where tmax is the maximizer of (6). From (8) above,

r is a monotonically decreasing function of tmax for
tmax 2 (0.5, 1). Hence tmax is a monotonic decreasing
function of r. h
Property 2. There exists a threshold rth such that
the decrement of r results in an increment of tmax
for r P rth, but tmax does not depends on r for
r 6 rth. The threshold rth is numerically 0.315.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

Proof. tF(t) is an increasing function for t 2

[0, 1]. tF(1  t) has a maximizer tmax 2 [0, 1] and
a minimum value 0 at t = 0 and 1. tF(1  t) increases for t 2 [0, tmax] and decreases for t 2 [tmax, 1]
monotonically. It follows from (6) that if
tmax 2 [0, 0.5], E(t) has a maximum at t = 0.5 and
if tmax 2 [0.5, 1), E(t) has
 a maximum at t = tmax.
rth is r such that E0 12 0. The numerical value
of rth is 0.315. h

1
0.9
0.8
0.7
Maximizer

2910

0.6
0.5
0.4
0.3

Figs. 810 depict how the objective function of

the proposed algorithm (MER) works. For these
gures, we assume that the selected node is located

0.2
0.1
0

GRS
MER with =0.01
MER with =0.03
MER with =0.05

0.8

Revenue

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

0.2

0.4
0.6
Normalized distance from a sender

0.8

Fig. 8. The revenue versus the distance from the sender when the
transmission range is 1.

1
0.9
0.8

Maximum Revenue

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

0.2
0.4
0.6
0.8
Normalized standard deviation of location errors

Fig. 10. The maximizer versus the standard deviation when the
transmission range is 1.

1
0.9

0.2
0.4
0.6
0.8
Normalized standard deviation of location errors

Fig. 9. The maximum revenue versus the standard deviation

normalized by the transmission range.

on OD in Fig. 6. Property 1 shows that MER works

identically to GRS except in the outskirts of the
transmission range, as shown in Fig. 8. However,
MER may require more hops to route packets from
source to destination compared with the case of perfect location environment, since the position of the
most likely chosen node, tmax, decreases due to location error, as in Figs. 8 and 10.
Property 2 gives us some insights on geographic
routing in an environment with location errors. If
the standard deviation of the location error is
greater than some threshold, i.e. r P rth, the maximizer of (6), tmax, decreases when r increases. If
r 6 rth, tmax does not depend on r, as in Fig. 10
and only the maximum revenue decreases as in
Fig. 9. tmax is xed at 0.5 after r = 0.315. In this
region, increasing r does not aect the selected
node, but decreases the revenue. This means that
for a standard deviation of error larger than rth,
taking error estimates into account does not help
in improving the performance of geographic routing. The algorithm is simply reduced to selecting
the node that is closest to the middle point of the
forward region. Hence, rth becomes the critical
point that determines the utility of the algorithm.
For example, when each node is equipped with a
GPS receiver that has a standard deviation of 3 m
of the location error, Bluetooth [29] which has a
transmission range of 10 m cannot use geographic
routing. However, a geographic routing scheme that
incorporates error information (such as MER) is
applicable in the case of IEEE 802.11 with a nominal transmission range 250 m.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

4. Simulation results
In this section, we use numerical simulations to
verify the performance of the proposed algorithm,
MER, in a mobile environment as in [11,18,20].
We use the following random way point (RWP)
mobility model. Each node chooses a destination
in a given area and moves at a constant speed,
which is uniformly chosen between 0 and 50 m/s.
The node stays for a pause time, which is uniformly
distributed between 0 s and 30 s. Each node broadcasts its own location periodically and proactively.
To avoid collisions, the interval of these broadcasts
is uniformly chosen from 1 s to 3 s, as in [18]. We
also set the neighbor timeout interval at 9 s. Since
location information errors in the mobile environment can be categorized into two classes, displacement due to mobility and location errors in
measurement, we use two cases of simulations in
the mobile environment in order to study the impact
of location errors on geographic routing. First, we
assume that location information errors exist only
in measurement. The displacements of the nodes
are assumed to be perfectly estimated in the mobile
environment. In the case when there is no location
error due to measurement, the authors in [20] study
the performance of geographic routing due to errors
in displacement using extensive simulations with
several mobility models. In our simulation environment, we compare the performance of MER versus
GRS when the parameters of the location error
model are known in Section 4.1. We further demonstrate the robustness of the algorithm to changing
parameters in the error model in Section 4.2.
Finally, we study the performance of MER in a
noisy environment due to inaccurate measurement
and mobility in Section 4.3. In order to reduce the
impact of mobility on the routing performance, we
adopt a displacement estimation method used in
[20,27]. Through our simulations we assume that
there is no time delay to route data from a source
to destination.
In our simulations we compare two dierent
schemes: GRS and MER. We use the performance
of GRS with perfect location information as an
upper bound on the routing performance of all
schemes in the presence of location errors.
To compare the fundamental performance in
Section 4.1, we use two measures: delivery success
rate and the number of attempts. In the other sections we compare delivery success rates. For the
delivery success rate, we do not use retransmissions

2911

or algorithms to nd alternative routes when packet

forwarding failures occur. Packet delivery is said to
succeed only if the packet is delivered to the destination by the schemes. For the number of attempts, we
employ an algorithm to nd another node when
packet forwarding fails. When the chosen node is
not available, a sender immediately nds the next
available node without retransmitting to the same
node. We dene this as a reattempt. When there is
no available node in the forward area, the routine
stops nding an alternative node. In the case when
there is no reattempt, the total number of attempts
is equivalent to the number of hops. The reattempts
require extra energy as well as time delay so that the
throughput will be degraded. Since the time delay
per reattempt is protocol-specic (such as the number of retransmission before declaring the chosen
node to be unavailable), hence we focus only on
the number of reattempts. We count the number
of attempts and reattempts until a hundred packets
are successfully delivered to their own destinations.
In MER, we use the distance from the maximizer
of the objective function to the transmission range
edge as a tuning parameter di. In order to study
how system parameters aect the performance of
geographic routing in the presence of location
errors, for our numerical results, we focus on the
case when the wireless environment is homogeneous
across all nodes. In practice, the wireless environment at each node in the network may be dierent.
For instance, in the case of GPS, each node could
receive a dierent number of satellite signals due
to obstacles. Our algorithm described in Section 3
can handle such a heterogeneous environment,
and we show that our algorithm works well in the
heterogeneous environment by simulations (Section
4.1).
We run 20 simulations with dierent random
seeds for each scenario and average the results.
4.1. The performance of MER when the distribution
of the location error is known
We investigate three scenarios, as illustrated in
Table 1. In the rst scenario, we deploy 100 nodes
in an area of 1000 1000 m2. The standard deviation of the location error for each node is 10 m.
We vary the transmission range from 25 m to
500 m. Fig. 11 shows that the transmission success
rate of MER is close to that of GRS with perfect
location information. However, the performance
of GRS degrades severely in the presence of location

2912

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

Table 1
Scenarios for simulations: A, N, R and r represent a deployed
area, the number of deployed nodes, the transmission range of
nodes and the standard deviation of location errors, respectively
Scenario

A (m2)

R (m)

r (m)

1
2
3

1000 1000
1000 1000
1000 1000

100
100
25500

25500
250
250

10 (240%)
350 (1.220%)
5 (2%)

100
90
80

Success rate [%]

70
60
50
40
30

tion. When we compare the total number of

attempts of MER and GRS in the presence of location errors, they are similar so that the reattempt
routine seems to dilute the benets of MER. However, GRS requires many more reattempts than
our algorithm. The reattempts may induce time
delay in the system. When the transmission ranges
increase, the number of attempts of GRS with location errors decreases since the failure probability
decreases, as in (1).
In the second scenario, the transmission range is
xed at 250 m, which is the nominal transmission
range of IEEE 802.11, and we vary the standard
deviation of the location error from 3 m (1.2%) to
50 m (20%). As expected, Fig. 12 shows that MER
performs much better than GRS when there are
location errors. The performance of GRS starts to
degrade when the standard deviation is above 3 m
(1.2%). However, the performance of MER does
not decrease signicantly. As the error increases,

20
GRS without error
MER with error
GRS with error

10
0

100
0

100

200
300
Transmission range

400

500

90

errors. Table 2 provides a comparison between

MER and GRS in terms of the total number of
attempts required for 100 successful transmissions.
Table 2 shows that MER with location errors needs
additional attempts (or hops) to reach its destination when compared with GRS with perfect location
information. The reason is that the location errors
cause MER to reduce its eective search range when
forwarding packets to its neighbors, as shown by
Property 1. Hence, MER requires additional
attempts to deliver packets from source to destina-

Success rate [%]

Fig. 11. The performance comparison of the forwarding schemes

in Scenario 1.

80

70

60

50
MER with error
GRS with error
40

5
10
15
Standard deviation of location errors[%]

20

Fig. 12. The performance comparison of the forwarding schemes

in Scenario 2.

Table 2
The comparison of the number of attempts in Scenario 1
ri (m)

GRS

MER

No error

300
350
400
500

Error

Error

(a)

(b)

(c)

(a)

(b)

(c)

(a)

(b)

(c)

(d)

238.15
208.50
184.95
155.45

0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0

260.00
225.65
195.95
162.05

17.85
14.75
9.40
5.70

0.0
0.0
0.0
0.0

255.70
219.05
193.45
159.85

1.20
0.60
0.50
0.10

0.0
0.0
0.0
0.0

107.37
105.06
104.60
102.83

(a) The average number of total attempts to deliver packets, (b) the average number of reattempts to nd alternative nodes, (c) the average
MER with error
number of times backward progress is made, (d) aaofofGRS
with no error %.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

2913

Table 3
The comparison of the number of attempts in Scenario 2
r (m)

GRS

MER

No error

3
5
7
10

Error

Error

(a)

(b)

(c)

(a)

(b)

(c)

(a)

(b)

(c)

(d)

289.75
289.75
289.75
289.75

0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0

295.95
300.35
303.05
310.60

5.30
8.55
11.15
17.05

0.00
0.00
0.05
0.10

295.45
300.10
304.35
310.35

0.10
0.65
1.20
1.35

0.00
0.00
0.05
0.10

101.97
103.57
105.04
107.11

(a) The average number of total attempts to deliver packets, (b) the average number of reattempts to nd alternative nodes, (c) the average
MER with error
number of times backward progress is made, (d) aaofofGRS
with no error %.

the eective search range in MER is further

reduced. This reduction in the eective search range
decreases the number of neighbor nodes to be
selected and degrades the performance. Similarly
to Scenario 1, MER with location errors requires
additional attempts to route packets when compared to the case with perfect information, as shown
in Table 3. When we compare GRS with MER,
GRS needs more reattempts than our algorithm in
the erroneous environment. Since MER and GRS
tend to choose the closest node to the transmission
range, they are less susceptible to backward progress. As analyzed in Section 2, backward progress
rarely happens, however, backward progress
increases with larger location errors.
In the third scenario, we vary the number of
deployed nodes from 25 to 500. We x the transmission range at 250 m and the standard deviation of
the location error at 5 m. Fig. 13 shows that the
performance MER is not aected by the number

100
95
90

Success rate [%]

85
80
75

of nodes while the performance of GRS is. The larger density reduces the distance between two adjacent nodes. This reduction in distance means that
the selected node is closer to the edge of the transmission range. Hence, transmission failure is more
likely to happen in GRS. Similarly to Scenario 1,
MER with location errors needs additional
attempts (Table 4) compared to the case without
location errors, however the success rate does not
decrease when the node density of the network is
increased. Note that in Table 4, unlike previous scenarios, the ratio of the additional attempts for
MER does not change signicantly as the number
of nodes increases, since the maximizer of the objective function depends only on the ratio of the location error to the transmission range. In the case of
GRS with errors, the total number of attempts
decreases as the number of nodes, N, increases
between 100 and 200. However, after that region,
the total number of attempts increases due to the
increment of the failure probability as the number
of nodes increases.
Depending on RF environments, the location
errors in the same system may be dierent. We use
the same settings as Scenario 3 except that now
there exits a shadowing area (250 250 m2) in which
nodes have twice the location errors of the other
areas. Fig. 14 shows that our algorithm still works
well in such a heterogeneous environment.
4.2. Robustness to estimation error

70
65
60
GRS without error
MER with error
GRS with error

55
50

100

200
300
The number of deployed nodes

400

500

Fig. 13. The performance comparison of the forwarding schemes

in Scenario 3.

In Section 2, we modeled the location error by a

Gaussian error distribution. In practice, the error
may not follow a Gaussian distribution and/or the
parameters of the model may be incorrect. In this
subsection, we study the robustness of MER with
respect to these two kinds of modeling errors: the
distribution function and parameter error.

2914

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

Table 4
The comparison of the number of attempts in Scenario 3
N

GRS

MER

No error

100
200
400
500

Error

Error

(a)

(b)

(c)

(a)

(b)

(c)

(a)

(b)

(c)

(d)

289.75
275.95
271.05
268.95

0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0

300.35
291.55
293.15
294.60

8.55
13.75
19.75
23.45

0.0
0.0
0.0
0.0

300.10
285.90
281.65
279.85

0.65
0.45
0.45
0.25

0.0
0.0
0.0
0.0

103.57
103.61
103.91
104.05

100

100

95

95

90

90

85

85
Success rate [%]

Success rate [%]

(a) The average number of total attempts to deliver packets, (b) the average number of reattempts to nd alternative nodes, (c) the average
MER with error
number of times backward progress is made, (d) aaofofGRS
with no error %.

80
75
70

75
70
GRS without error
MER with Gaussain error
MER with uniform error
MER with exponential error
GRS with Gaussain error
GRS with uniform error
GRS with exponential error

65

65

60

60
GRS without error
MER with error
GRS with error

55

55
50

100

200
300
The number of deployed nodes

400

500

Fig. 14. The performance comparison of the forwarding schemes

in Scenario 3 with a heterogeneous RF environment.

In Fig. 15, we simulate three dierent location

error models: uniformly distributed error, exponentially distributed error, and Gaussian error. However, the MER algorithm always assumes a
Gaussian model. The transmission range of each
node is 250 m and the standard deviation of the
location error is 5 m. The simulation results show
that MER is quite robust to dierence in error distribution functions.
In Fig. 16, the underlying error model is also
Gaussian. However the parameter used by MER
is dierent from the true parameter of the underlying model. In the simulation the transmission range
of each node is 250 m and the standard deviation of
the location error is assumed to be 5 m. However,
the actual standard deviation of the location error
is varied from 0 m to 10 m. Fig. 16 illustrates that
MER is robust to the estimation error and outperforms GRS.

100

200
300
The number of deployed nodes

400

500

Fig. 15. The performance of forwarding schemes versus the

number of nodes with dierent error distributions and the xed
transmission range.

100
98
96
94
Success rate [%]

50

80

92
90
88
86
84
82
80

MER
GRS
0

4
6
Real Location Error

10

Fig. 16. The performance of forwarding schemes versus the

parameter error in estimation with the xed number of nodes and
the xed transmission range.

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

4.3. Noisy mobile environment

100

X 0 X 1

X X
 t0  t1 ;
t1  t2

where X(0) is the predicted position at the current

time t(0), X(1) is the reported location at the rst recent time t(1), and X(2) is the reported location at the
second recent time t(2). The method improves the
routing performance by reducing the mobility error
when the reported location information is assumed
to be perfect. However, this method will accumulate
measurement and prediction errors when reported
location information is noisy. Hence, we use the instant velocity, which is available to GPS equipped
nodes, since the velocity measured by GPS is considered to be very accurate [6]. The instant velocity
is announced with location information. Each node
predicts neighbors positions when forwarding a
packet as follows.
X 0 X 1 v1  t0  t1 ;
(0)

10

where X is the predicted position at current time

t(0), X(1) is the reported location at the rst recent
time t(1), and v(1) is the reported velocity at t(1).
Fig. 17 shows the simulation results when the
location error at each node has a Gaussian distribution with standard deviation 5 m and the transmission range is xed at 250 m. GRS with the
displacement prediction performs better than GRS
without the displacement prediction, as shown in
[20]. MER is slightly aected by mobility but outperforms GRS in all cases. As shown in [11], the
performance degradation from mobility can be further improved by reducing time discrepancy. However, the measurement error of X(1) in (10) does
not decrease, as the time discrepancy reduces. The
reduction of the time discrepancy also does not
aect the location error of a sender. Hence, any
eort to reduce the time discrepancy from mobility
cannot help mitigate the impact of measurement
errors on geographic routing.

95
90
Success rate [%]

In this subsection, we compare the performance

of MER and GRS in the presence of location errors
due to inaccurate measurement and mobility. When
each node chooses the next node, we use the predicted positions of neighbors at the transmission
time similar to [20,27] in order to improve the routing performance. The authors in [20,27] predict the
neighbor positions at the transmission time by using
two positions reported at two recent times as
follows:
1

2915

85
80
75
70
65
60
100

MER with perfect prediction

MER with prediction
MER without prediction
GRS with perfect prediction
GRS with prediction
GRS without prediction
150

200
250
300
350
400
450
The number of nodes in 1000 by 1000 square area

500

Fig. 17. The performance of forwarding schemes versus the

number of nodes with Gaussian location errors and the RWP
mobility model.

4.4. Discussion
In order to solve the fundamental problem of
geographic routing with location errors, in this
paper we do not use a protocol specic solution to
help mitigate errors. However, such a protocol-specic solution could in certain cases help combat
location errors [30]. For example, in the case when
nodes are static and have xed transmission ranges,
the transmission failures caused by asymmetric
communication links can be reduced by a threeway handshake protocol when nodes join the network. However, the three-way communication does
not alleviate backward progress. Even in a static
wireless network, such a protocol cannot avoid
transmission failures when the transmission range
is controllable in the presence of location errors.
Further, the solution is not suitable in a dense
mobile wireless network, where frequent topology
changes may take place. Moreover, in contrast to
our proactive announcement (one-way communication) of location information, the three-way communication results in 2n(n  1) overhead messages
per time interval in a transmission range, where n
represents the number of nodes within a transmission range. The excessive overhead messages may
in fact worsen network performance such as
throughput. For the above mentioned reasons, in
this paper, we do not focus on protocol-specic
solutions to alleviate location errors. However, such
approaches can be potentially used in conjunction
with our method on a case by case basis.

2916

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

Another feature of our approach is to provide an

understanding of the intrinsic performance achievable using geographic routing with location errors.
This serves to provide design guidelines for implementers to choose equipment that provides an
appropriate level of accuracy.
5. Conclusions
In this paper, we consider the impact of location
errors on geographic routing in multi-hop wireless
networks. We have shown that location errors can
signicantly impact the performance of geographic
routing. The degradation in the routing performance
depends on the transmission range of the sender,
error characteristics of the sender and its neighbors,
and the deployed density of nodes. We have proposed
a new algorithm called MER in order to mitigate the
eect of noisy location information by explicitly considering the error probability when making routing
decisions. In doing so we nd that our algorithm performs quite well, in many cases, even approximating
the performance of geographic routing without location errors. We have also used simulations to show
that MER is robust to dierent location error models
and errors in model parameters.
References
[1] F. Kuhn, R. Wattenhofer, A. Zollinger, Worst-case optimal
and average-case ecient geometric ad-hoc routing, in:
ACM MobiHoc03, June 2003, pp. 267278.
[2] M. Mauve, J. Widmer, H. Hartenstein, A survey on positionbased routing in mobile ad hoc networks, IEEE Network 15
(2001) 3039.
[3] S. Ratnasamy, B. Karp, L. Yin, F. Yu, D. Estrin, R.
Govindan, S. Shenker, GHTa geographic hash table for
data-centric storage, in: First ACM International Workshop
on Wireless Sensor Networks and their Applications, 2002,
pp. 7887.
[4] I. Stojmenovic, Position-based routing in ad hoc networks,
IEEE Communications Magazine 40 (7) (2002) 28.
[5] P. Misra, B.P. Burke, M.M. Pratt, GPS performance in
navigation, Proceedings of the IEEE 87 (1) (1999) 6585.
[6] P. Misra, P. Enge, Global Positioning System: Signals,
Measurements, and Performance, Ganga-Jamuna Press,
Massachusetts, 2001.
[7] P. Bahl, V.N. Padmanabhan, RADAR: an in-building RFbased user location and tracking system, in: IEEE INFOCOM00, March 2000, pp. 775784.
[8] N. Bulusu, J. Heidemann, D. Estrin, GPS-less low-cost
outdoor localization for very small devices, IEEE Personal
Communications (October) (2000) 2834.
[9] L. Doherty, K.S.J. Pister, L.E. Ghaoui, Convex position
estimation in wireless sensor networks, in: IEEE INFOCOM01, April 2001, pp. 16551663.

[10] J. Hightower, G. Borriello, Location systems for ubiquitous

computing, Computer 34 (8) (2001) 5766.
[11] B. Karp, Geographic routing for wireless networks, Ph.D.
thesis, Harvard University, 2000.
[12] H. Takagi, L. Kleinrock, Optimal transmission range for
randomly distributed packet radio terminals, IEEE Transactions on Communications 32 (3) (1984) 246257.
[13] G. Finn, Routing and addressing problems in large metropolitan-scale internetworks, Technical Report ISI/RR-87180, USC/ISI, March 1987.
[14] T.C. Hou, V.O.K. Li, Transmission range control in
multihop packet radio networks, IEEE Transactions on
Communications 34 (1) (1986) 3844.
[15] E. Kranakis, H. Singh, J. Urrutia, Compass routing on
geometric networks, in: The 11th Canadian Conference on
Computational Geometry, Vancouver, Canada, August
1999.
[16] I. Stojmenovic, X. Lin, Loop-free hybrid single-path/ooding routing algorithms with guaranteed delivery for wireless
networks, IEEE Transactions on Parallel and Distributed
Systems 12 (10) (2001) 246257.
[17] P. Bose, P. Morin, I. Stojmenovic, J. Urrutia, Routing with
guaranteed delivery in ad hoc wireless networks, ACM
Wireless Networks 7 (6) (2001) 609616.
[18] B. Karp, H.T. Kung, GPSR: greedy perimeter stateless
routing for wireless networks, in: ACM MobiCom00,
August 2000, pp. 243254.
[19] Y.B. Ko, N.H. Vaidya, Location-aided routing (LAR) in
mobile ad hoc networks, ACM/Baltzer Wireless Networks
(WINET) Journal 6 (4) (2000) 307321.
[20] D. Son, A. Helmy, B. Krishnamachari, The eect of
mobility-induced location errors on geographic routing in
ad hoc networks: analysis and improvement using mobility
prediction, in: IEEE Wireless Communications and Networking Conference (WCNC04), March 2004.
[21] S. Slijepcevic, S. Megerian, M. Potkonjak, Location errors in
wireless embedded sensor networks: sources, models, and
eects on applications, ACM SIGMOBILE Mobile Computing and Communications Review 6 (3) (2002) 6778.
[22] Y. Kim, J. Lee, A. Helmy, Impact of location inconsistencies
on geographic routing in wireless networks, in: ACM
International Workshop on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM03), July
2003, pp. 124127.
[23] Garmin Ltd. Available from: <http://www.garmin.com>.
[24] Trimble Navigation Ltd. Available from: <http://www.
trimble.com>.
[25] R. Durrett, Probability: Theory and Examples, second ed.,
Duxbury Press, Belmont, CA, 1996.
[26] T. He, C. Huang, B.M. Blum, J.A. Stankovic, T. Abdelzaher, Range-free localization schemes for large scale sensor
network, in: ACM MobiCom03, September 2003, pp. 81
95.
[27] I. Stojmenovic, M. Russell, B. Vukojevic, Depth rst search
and location based localized routing and qos routing in
wireless networks, Computers and Informatics 21 (2) (2002)
149165.
[28] A. Paoulis, S.U. Pillai, Probability, Random Variables, and
Stochastic Process, fourth ed., McGraw-Hill, New York,
2002.
[29] A. Davies, An overview of bluetooth wireless technology and
some competing lan standards, in: The 1st IEEE Interna-

S. Kwon, N.B. Shro / Computer Networks 50 (2006) 29022917

tional Conference on Circuits and Systems for Communications, St. Petersburg, Russia, June 2002, pp. 206211.
[30] M. Zorzi, R.R. Rao, Geographic routing forwarding
(GeRaF) for ad hoc and sensor networks: multihop performance, IEEE Transactions on Mobile Computing 2 (4)
(2003) 337348.

Sungoh Kwon received his B.S. and M.S.

degrees in electrical engineering from
KAIST, Daejeon, Korea, in 1994 and
1996, respectively. From 1996 to 2001,
he had worked as a research sta in
Shinsegi Telecomm Inc., Seoul, Korea.
Since 2001, he has worked for the Ph.D.
in the School of Electrical and Computer
Engineering from Purdue University. His
current interests are in wireless communication networks.

Ness B. Shro is a Professor of Electrical

and Computer Engineering at Purdue
and director of the Center for Wireless
Systems and Applications (CWSA), a
university-wide center on wireless systems and applications. His research
interests span the areas of wireless and
wireline communication networks, where
he investigates fundamental problems in
the design, performance, pricing, and
security of these networks.

2917

He is an active member in the networking research community

and has been involved in the leadership of various conferences
and journals. He currently serves on the editorial boards of
IEEE/ACM Transactions on Networking and the Computer
Networks Journal. He was the technical program co-chair of
IEEE INFOCOM03, the premier conference in communication
networking. He was also the conference chair of the 14th Annual
IEEE Computer Communications Workshop (CCW99), the
program co-chair for the symposium on high-speed networks,
IEEE Globecom 2001, and the panel co-chair for ACM Mobicom02. He was also a co-organizer of the NSF workshop on
Fundamental Research in Networking, held in Arlie House
Virginia, in 2003.
He has received numerous awards for his networking research,
including the NSF CAREER award, the best paper of the year
award for the Computer Networks journal, and the best paper of
the year award for the Journal of Communications and Networks
(JCN).