Fusion of Inertial Sensors and OFDM Signals of

Opportunity for Unassisted Navigation

Jason Crosby, Richard Martin, John Raquet, and Michael Veth
AbstractThe advent of the global positioning system
(GPS) has provided worldwide high-accuracy position
measurements. However, GPS may be rendered unavail-
able by jamming, disruption of satellites, or simply by
signal shadowing in urban environments. Thus, this paper
considers fusion of Inertial Navigation Systems (INS)
and Orthogonal Frequency Division Multiplexed (OFDM)
Signals of Opportunity (SoOP) for navigation. Typical
signal of opportunity navigation involves the use of a
reference receiver and uses time difference of arrival
(TDOA) measurements. However, by exploiting the block
structure of OFDM communication signals the need for
the reference receiver is removed.
Index Termsinertial navigation, OFDM, signals of
opportunity.
I. INTRODUCTION
One alternative method of navigation that has
been proposed recently is Navigation via Signals
of Opportunity. The idea is to use existing radio
frequency signals that were not intended for nav-
igation, and leverage knowledge of the transmitter
locations to determine the position of the receiver.
Some signals investigated thus far include the Na-
tional Television System Committee Broadcast Sig-
nal, AM and FM signals [1,2,3]. This problem is the
converse of the source localization problem, hence
similar mathematical methods can be used to extract
a position estimate.
Measurements that can be taken to aid in position-
ing include the angle of arrival (AOA), the received
signal strength (RSS), or the time difference of
arrival (TDOA) at multiple receivers. We focus on
TDOA, since it is difcult to get sufcient position
accuracy from AOA or RSS [4]. One difculty
encountered with using TDOA is that there must
be either two transmitters sending the same signal
or two spatially separated receivers measuring the
same transmission. Usually only one transmitter is
available, hence a reference receiver must be placed
at a known location, and the mobile (whose posi-
tion is to be determined) must cooperate with the
reference in some way. A second difculty is that
TDOA measurements generally require some form
of correlation between the two received signals. This
requires that the reference retransmit much of its
received signal to the mobile, using signicant band-
width. Recent previous work has shown how OFDM
modulation can alleviate the bandwidth constraint
[5].
By using the same blind block synchronization
described later a received time for each OFDM
block can be obtained for both a reference and
mobile receiver. After individual blocks are dis-
tinguished, statistical features such as mean and
standard deviation of each block can be calculated.
These statistical features along with received times
for the corresponding block are then transmitted
from the reference receiver to the mobile receiver
instead of transmitting the entire signal, greatly
reducing required bandwidth. These statistical fea-
tures are then correlated with statistical features
calculated at the mobile receiver. Correlating the
features allows a time difference to be obtained [5].
In this paper, we leverage the previous work
to show how OFDM signals of opportunity can
be combined with an inertial navigation system
(INS) to remove the need for a reference receiver
almost entirely, with the almost caveat remaining
because a reference receiver is still occasionally
needed to account for transmitter clock drift. This
TDOA measurement system is shown in Figure 1
for a single transmitter, though in general multiple
transmitters would be used.
INS systems are strong candidates for non-GPS
navigation because they have high accuracy over
short periods of time. However, they are subject to
gradual drift. The use of a signal of opportunity
can mitigate this drift. Specically, we show that an
Fig. 1. TDOA geometry. The TDOA is computed by calculating
block arrival times at the initial location, and at the nal location.
The time difference can be obtained by differencing these two
measurements.
INS can be used to obtain multiple relative position
estimates at relative times. Then the proposed fu-
sion process can compare the expected and actual
times of reception of the OFDM systems block
structure to extract TDOA information. Specically,
the receiver can locate the start of each OFDM
block in blind fashion by looking for the repetition
induced by the cyclic prex of OFDM systems. The
receiver can compare the measured reception times
of sequential blocks to the known block length,
and a change in the expected vs. measured time
indicates if the receiver is moving towards or away
from a given transmitter. With enough transmitters,
a position track can be obtained.
II. INS SYSTEM MODEL
The INS model used in this research was a simple
two dimensional model. Two accelerometers (f
xb
and f
yb
) and one gyroscope (

) were used to obtain
acceleration and angular rotation rate in a plane.
Additive white Gaussian noise (AWGN) was added
to the accelerometer and gyroscope measurements.
AWGN added to these measurements had a standard
deviation () of 0.1
m
s
2
for the accelerometer and
0.1
rad
s
for the gyroscope. One accelerometer started
pointing in the positive X direction, and the other
in the positive Y direction. This system is shown in
Figure 2. The body accelerations were resolved in
the X and Y directions using the formulas found in
[6] which are:
Fig. 2. INS model used in the simulation. A two dimensional model
with two accelerometers and one gyroscope. Position is obtained in
the X and Y directions
f
xi
= f
xb
cos() +f
yb
sin() (1)
f
yi
= f
xb
sin() +f
yb
cos() (2)
where f
xb
and f
yb
are measured acceleration in
the X
b
and Y
b
body frame. is the rotation
angle obtained by integrating

. f
xi
and f
yi
are
accelerations resolved in the X and Y directions.
INS measurements were taken at 50Hz. X and Y
position were then obtained via the formulas:
v
xi
=

f
xi
(3)
v
yi
=

f
yi
(4)
X =

v
xi
(5)
Y =

v
yi
(6)
Although the AWGN added to the simulated
measurements was zero mean, these errors cause
an error drift due to the fact that they are integrated
twice. This causes errors in INS position solution
to grow over time. Because of these continuously
growing errors, INS systems are typically aided by
some sort of measurement system. Here we will
use OFDM signals of opportunity to aid the INS
system through the use of a Kalman lter. Truth
data was obtained by integrating the simulated INS
measurements without adding AWGN.
III. OFDM SYSTEM MODEL
The rst step in an OFDM communication system
occurs at the bit level. Data bits to be transmitted
go through an interleaving process and are then
encoded. The use of a convolutional encoder is
common, but not necessarily required. The encoding
rate may also be varied depending on transmission
bit rate requirement. This serial bit stream is then
converted into a parallel bit stream. These paral-
lel bits are then mapped to a signal constellation
which depends on modulation type used by the
transmitter. Modulation type can vary depending on
transmission bit rate requirement, but typically in-
clude BPSK, QPSK, 16 QAM, and 64 QAM. These
constellation mappings are then applied to numerous
carriers. This simulation included N = 64 carriers.
Then an N point Inverse Fast Fourier Transform
(IFFT) is performed on the data modulated carriers
to get the data from the frequency domain to the
time domain. This action produces N time domain
samples. These N samples are then demultiplexed
from parallel to serial. Once in serial a cyclic prex
is added to the beginning of the signal. This is done
by copying the last v time domain samples and
appending them to the front of the original N time
domain samples. This creates an M = N+v sample
OFDM symbol. This symbol is then modulated at
the carrier frequency and transmitted over the air. A
block diagram of an OFDM communication system
is shown in Figure 3.
The cyclic prex (CP) is of particular importance
to these experiments because it is the sole piece
that allows for the removal of the reference receiver.
The addition of the CP was originally designed to
mitigate the effects of multipath, specically inter-
symbol interference (ISI) and intercarrier interfer-
ence (ICI). The use of the CP removes these effects
[7]. Because the CP is an exact replica of the last
section of the symbol, an autocorrelation algorithm
will produce a well dened peak at the point when
these two sections of the symbol are correlated.
This peak becomes even more distinguishable when
several symbol peaks are averaged. Finding the
index of where this peak occurs allows for a TDOA
measurement to be obtained without the need for a
reference receiver. This process will be explained in
the next section.
Fig. 3. OFDM communication block diagram. After coding and
interleaving, bits to be transmitted are mapped using a signal constel-
lation onto multiple carrier frequencies. An IFFT is then performed
on the carrier frequencies. Once in the time domain a cyclic prex
is added and the symbol is transmitted over the channel.
IV. TDOA COMPUTATION
In this section we discuss how TDOA measure-
ments are obtained from the OFDM signals. To do
this a receiver estimates block boundaries within the
signal. This is a common method of blind block
synchronization and is derived in [8]. Thus given
a received signal the maximum likelihood (ML)
estimate of the block boundaries is

ML,rx
= arg max
0mM1
{
avg
(m)} (7)
where

avg
(m) =
K1

k=0
m+v

i=m+1
y
rx
(Mk +i)y

rx
(Mk +i +N)
(8)
and is the real operator, K is the number of blocks
averaged over, M is the symbol length, k is the
index of the OFDM block, and i is the index of the
sample within the OFDM block.
Finally, once

ML,rx
is obtained the TDOA can
be calculated using the formula
TDOA = (

ML,rx

initial
) T
s
(9)
where T
s
is the sampling interval. Note that in these
simulations
initial
is always zero, and all TDOA
measurements are measured from the initial point
(0,0).
V. SIMULATION RESULTS
This section illustrates performance of the TDOA
aided INS system. Note that no transmitter or re-
ceiver clock errors were modelled in the simulation.
For the simulations the OFDM signal used N = 64
carriers with a cyclic prex length of v = 16.
This produced an M = 80 sample OFDM symbol.
These lengths are consistent with the IEEE 802.11
0 100 200 300 400 500
50
0
50
100
150
200
X position (m)
Y

p
o
s
i
t
i
o
n

(
m
)
X vs Y position of TDOA system with 3 Tx


Actual Position
INS Only Position
TDOA Aided Position
Fig. 4. TDOA aided INS position. Three transmitters were used to
obtain TDOA measurements and aid the INS through the use of a
Kalman lter.
0 100 200 300 400 500
50
0
50
100
150
200
X position (m)
Y

p
o
s
i
t
i
o
n

(
m
)
X vs Y position of TDOA system with 3 oversampled Tx


Actual Position
INS Only Position
TDOA Aided Position
Fig. 5. 15 oversampled TDOA aided INS position. Again
three transmitters were used to obtain TDOA measurements. These
measurements used oversampling of the OFDM signal producing
more accurate measurements.
standard. The transmitter locations were assumed to
be known in relation to the receiver, and the receiver
position was initialized to (0,0). Three transmitters
were used and stationed in a triangular fashion
around the receiver. Once INS measurements and
TDOA measurements had been simulated using
MATLAB, a Kalman Filter also implemented in
MATLAB was used to output an optimal position
estimate. Figure 4 shows the position result from the
output of the Kalman lter. It can be seen that after
a short period, the aided system performs better than
0 20 40 60 80 100 120
10
5
0
5
10
time (s)
X

e
r
r
o
r
s

(
m
)
Errors with 1 sigma bounds for 3 TX


X error
+ 1 sigma bound
0 20 40 60 80 100 120
10
5
0
5
10
time (s)
Y

e
r
r
o
r
s

(
m
)


Y error
+ 1 sigma bound
Fig. 6. Errors with 1 error bounds. Error bounds shown for X and
Y position are used to ensure the Kalman lter is operating properly.
0 20 40 60 80 100 120
1
0.5
0
0.5
1
time (s)
X

e
r
r
o
r
s

(
m
)
Errors with 1 sigma bounds for 3 oversampled TX


X error
+ 1 sigma bound
0 20 40 60 80 100 120
1
0.5
0
0.5
1
time (s)
Y

e
r
r
o
r
s

(
m
)


Y error
+ 1 sigma bound
Fig. 7. Oversampled errors with 1 error bounds. Error bounds
shown for X and Y position are used to ensure the Kalman lter is
operating properly. Error bounds from the oversampled system are
smaller than error bounds of the regular system.
the INS alone.
This simulation was rerun using an oversampled
TDOA system. This system used a sampling rate
15 greater than the original system. Figure 5
shows the position result from the output of the
Kalman Filter when the OFDM signal is oversam-
pled by 15 times. This oversampled system clearly
produces better position estimates than compared to
the rst simulation. Oversampling the system pro-
duces more sample measurements with independent
noise values. If there were any time correlated errors
in the TDOA measurement there would be less of
0 20 40 60 80 100 120
0
5
10
15
20
25
30
35
40
45
50
time (s)
R
M
S

e
r
r
o
r

(
m
)
RMS error vs time of TDOA aided system and INS only System with 3 Tx


TDOA aided errors
Oversampled TDOA aided errors
INS only errors
Fig. 8. RMS error with 3 transmitters. The above simulations were
run 10 times each and RMS errors were calculated. INS errors grow
over time, while the TDOA aided system are much lower.
0 100 200 300 400 500
50
0
50
100
150
200
250
X position (m)
Y

p
o
s
i
t
i
o
n

(
m
)
X vs Y position of TDOA system with 1 Tx


Actual Position
INS Only Position
TDOA Aided Position
Fig. 9. TDOA aided INS position. One transmitter was used to obtain
a TDOA measurement and aid the INS. When compared to Figure 4
this one measurement is less accurate than using three measurements.
an improvement.
The errors in the X and Y directions of the
two systems were also plotted along with 1 error
bounds. These plots are shown in Figures 6 and
7. Both simulations were then run ten times and
the average RMS error for each case over time
was computed and compared with the average RMS
error of the INS only system. Average RMS error
for all three systems are shown in Figure 8.
Simulations were also run using only 1 OFDM
transmitter. Results from the regular system are
shown in Figure 9 and results from the oversampled
0 100 200 300 400 500
50
0
50
100
150
200
X position (m)
Y

p
o
s
i
t
i
o
n

(
m
)
X vs Y position of oversampled TDOA system with 1 Tx


Actual Position
INS Only Position
TDOA Aided Position
Fig. 10. Oversampled TDOA aided INS position. Again only
one transmitter was used to obtain a TDOA measurement. This
measurement used oversampling of the OFDM signal producing
a more accurate measurement. When compared to Figure 9 this
oversampled measurement is more accurate than the non-oversampled
system.
0 20 40 60 80 100 120
40
20
0
20
40
time (s)
X

e
r
r
o
r
s

(
m
)
Errors with 1 sigma bounds for 1 TX


X error
+ 1 sigma bound
0 20 40 60 80 100 120
20
10
0
10
20
time (s)
Y

e
r
r
o
r
s

(
m
)


Y error
+ 1 sigma bound
Fig. 11. Errors with 1 error bounds. Error bounds shown for X and
Y position are used to ensure the Kalman lter is operating properly.
system are shown in Figure 10. Errors and error
bounds of the X and Y position are shown in
Figures 11 and 12 Once again, average RMS error
over 10 runs was compared with the INS only case.
These results are shown in Figure 13.
0 20 40 60 80 100 120
40
20
0
20
40
time (s)
X

e
r
r
o
r
s

(
m
)
Errors with 1 sigma bounds for 1 oversampled TX


X error
+ 1 sigma bound
0 20 40 60 80 100 120
20
10
0
10
20
time (s)
Y

e
r
r
o
r
s

(
m
)


Y error
+ 1 sigma bound
Fig. 12. Oversampled errors with 1 error bounds. Error bounds
shown for X and Y position are used to ensure the Kalman lter is
operating properly.
0 20 40 60 80 100 120
0
10
20
30
40
50
60
time (s)
R
M
S

e
r
r
o
r

(
m
)
RMS error vs time of TDOA aided system and INS only System with 1 Tx


TDOA aided errors
Oversampled TDOA aided errors
INS only errors
Fig. 13. RMS error with 1 transmitter. Systems used in Figures
9 and 10 were run 10 times and RMS error was calculated. When
compared to Figure 8 the RMS error is higher with only 1 transmitter.
VI. CONCLUSIONS AND FUTURE WORK
The use of OFDM modulated signals of oppor-
tunity has been shown as a viable option to aid in-
ertial navigation systems. Furthermore exploitation
of the block structure allowed the removal of the
reference receiver typically needed to calculate time
difference of arrival measurements. The accuracy of
these measurements has also been shown to improve
with the use of oversampling. This oversampling
causes smaller sampling intervals and thus a ner
TDOA measurement can be obtained. This more
accurate measurement fed into the Kalman lter
causes a more accurate position estimate. Overall
position accuracy also increases as the number of
transmitters increases. This is expected as more
measurements contain more information about the
receivers position.
The models used thus far have been greatly
simplied from that of an actual system. Future
work will consider effects of receiver clock errors
and clock drift, as well as transmitter clock errors
and clock drift which are known to cause signif-
icant errors in positioning. Future efforts will also
incorporate a rigorous 3D INS model, and use actual
INS measurements with differential GPS truth data
to simulate TDOA measurements.
ACKNOWLEDGMENT
This work was supported in part the the Air Force
Research Laboratory, Sensors Directorate and the
Air Force Ofce of Scientic Research. The views
expressed in this paper are those of the authors, and
do not reect the ofcial policy or position of the
United States Air Force, Department of Defense,
or the U.S. Government. This document has been
approved for public release; distribution unlimited.
REFERENCES
[1] Ryan J. Eggert, Evaluating the Navigation Potential of
the National Television System Committee Broadcast Sig-
nal, Masters thesis, Air Force Institute of Technology,
Wright-Patterson AFB OH, March 2004. [Online.] Available:
http://handle.dtic.mil/100.2/ADA423977.
[2] Brian S. Kim, Evaluating the Correlation Characteristics of
Arbitrary AM and FM Radio Signals for the Purpose of Naviga-
tion, Masters thesis, Air force Institute of Technology, Wright-
Patterson AFB OH, March 2006.
[3] Johnathan A. McEllroy, Navigation Using Signals of
Opportunity in the AM Transmission Band, Masters
thesis, Air Force Institute of Technology, Wright
Patterson AFB OH, September 2006. [Online.] Available:
http://handle.dtic.mil/100.2/ADA456511.
[4] James J. Caffery, Jr. and Gordon L. Stuber, Overview of Radi-
olocation in CDMA Cellular Systems, IEEE Communications
Magazine, vol. 36, no. 4, pp. 38-45, April 1998.
[5] Jamie S. Velotta, Navigation Using Orthogonal Frequency
division Multiplexed Signals of Opportunity, Masters thesis,
Air Force Instutute of Technology, Wright Patterson AFB OH,
September 2007.
[6] David Titterton and John Weston, Strapdown Inertial Naviga-
tion Technology, 2nd Edition, IEE Radar, Sonar, Navigation and
Avionics Series 17, pp. 19, 2004.
[7] Werner Henkel, Georg Taubock, Per Odlind, Per Ola Borjesson,
and Niklas Petersson, The Cyclic Prex of OFDM/DMT - An
Analysis, 2002 International Zurich Seminar on Broadband
Communications, Access, Transmission, Networking, 22.1-22.3,
IEEE, Zurich, Switzerland, February 2002.
[8] Jan-Jaap van de Beek, Magnus Sandell, and Per Ola Borjesson,
ML Estimation of Time and Frequency Offset in OFDM sys-
tems, IEEE Transactions on Signal Processing, vol. 45, no. 7,
pp. 1800-1805, July 1997.