Journal of Global Positioning Systems (2007)
Vol.6, No.2: 149-157
Autonomous Navigation Environment with Self-Calibrating
Transceivers
S. Schlötzer, S. Martin and M. v. Voithenberg
EADS Astrium, Germany
Abstract. An operable navigation system which
demonstrates successful self-calibration and precise local
navigation has been developed by EADS Astrium. This
paper presents the architecture of the autonomous
navigation environment with the ability to calibrate itself
as well as the results of field tests. The Self-calibrating
autonomous Navigation Environment (SekaN) can be
used as stand-alone navigation system for applications
where satellite signals are not available or where
autonomy and high precision is required. Cargo drop,
navigation in canyons and open pit mines, indoor
navigation and extraterrestrial navigation are just
examples of possible applications. The self-calibrating
feature of SekaN is of special interest in conflict areas
where a temporary autonomous navigation environment
has to be installed quickly and where it is not possible to
calibrate the locations of the pseudolites a priori.
Furthermore, the system can be operated as augmentation
system to classical satellite navigation systems. Therefore
a mixed mode has been introduced which allows for
simultaneous tracking of both satellite signals and
pseudolite signals. Referencing of the local coordinate
system to e.g. WGS84 becomes possible. The SekaN
system comprises the following HW units developed by
EADS Astrium: at least 4 Transceivers (TCs), a Rover
receiver (ROV) and a Master Control Station (MCS). A
WLAN data link is used between the units. Each TC
comprises a GNSS signal generator NSG 5100 which
supports both GPS and Galileo signals and an Astrium-
specific GPS/PSL receiver. The number of TCs in the
network is scalable and dependent on the specific
application of the SekaN. Various TC-array sizes are
supported as the output power of the pseudolites can be
varied in a wide range. The rover receiver positioning
takes place at the MCS. However, several receivers may
be registered at the MCS. The TCs are operated
unsynchronized and differential concepts are applied to
eliminate the clock errors. Presently the pulsed signals
with pseudolite spreading codes at the GPS L1 and
dummy navigation messages are used as navigation
signals. As soon as low-cost Galileo receivers are
available the system can be switched to any Galileo
frequency band. In a batch process the exact locations of
the TC TX-antennas are determined without any a priori
knowledge of the geometric array configuration. The
general idea behind the self-calibration algorithms is
based on the solution algorithm for self-calibrating
pseudolite arrays presented in (LeMaster and Rock,
2002). However, several modifications were necessary to
adapt the algorithms to the SekaN system requirements.
The rover which is used for data collection during the
self-calibration process is designed as a Receiver-only
module instead of a TC module. This makes the rover
hardware less complex, smaller and lighter, but also
complicates the self-calibration process. Self-differencing
between the stationary TCs and the rover TC can no
longer be applied. The ranges between the rover RX and
the TCs are therefore not directly observable. The self-
calibration and navigation algorithms developed for the
SekaN work in both 2-D and 3-D scenarios. Although
multipath effects, non-linearities and the near-far-effect
are inherent in these kinds of ground-based navigation
systems, precise user positioning at the sub-meter level
becomes possible even with low-cost receivers within the
self-calibrated navigation environment.
Keywords: Pseudolites, Autonomous Navigation, GPS,
Galileo
1. INTRODUCTION
In this research funded by the DLR (German Aeronautic
and Space Agency) a navigation environment called
SekaN with self-calibrating transceivers has been
developed. The research aims at the provision of an
operational system which develops new fields of
applications which are listed hereafter. For many
applications it is highly desirable to become independent
from the availability of GPS or Galileo signals.
Navigation in buildings, mines, urban canyons or on

150 Journal of Global Positioning Systems
remote planets suffers from very weak satellite signal
strengths or poor DOP values which make satellite
navigation almost impossible. For other applications it is
of interest to have an auxiliary system to satellite-based
systems in order to enhance the accuracy, reliability and
availability of the navigation solution. Safety-critical
applications can be listed like cargo drop landing,
precision approach and landing and the guidance of
mobile troops. Pseudolites can be set up in the desired
environment to improve the availability of navigation
signals. Inexpensive non-satellite based navigation
systems are restricted to relatively small operation areas
where good signal coverage is ensured by the pseudolites.
Therefore it is of special interest that the autonomous
navigation environment can be put into operation quickly
at different locations. The self-calibration feature of the
SekaN system speeds up the system start up and provides
a navigation environment where it is not possible to
determine the pseudolite locations a priori through
geodetic survey. Pseudolite-only navigation as well as
navigation with both satellite-signals and pseudolite-
signals is possible in the SekaN system. Even low-cost
off-the-shelf receivers can be used in the autonomous
navigation environment with a minimum effort of
adaptation.
In the following sections the SekaN system hardware and
software is presented as well as the results of field tests.
Reliable self-calibration of the TC network and carrier
phase based user positioning relying on pseudolite-only
signals are demonstrated.
2. SYSTEM ARCHITECTURE
The SekaN prototype system comprises 6 Transceivers
(TCs), a Master and Control Station (MCS) and a Rover
receiver (ROV). An extension to more than 6 TCs and to
more than one rover RX is possible without any software
modifications.
Transceiver Architecture
The TC architecture comprises a GNSS signal generator,
a GPS/PSL receiver, a WLAN access point client and a
common power unit. All components are integrated in a
common 19” TC rack which is presented in Fig. 1.
Two separate antennas are used for the signal emission
and the signal reception. The RX antenna is mounted on
top of the TX antenna at an adequate spacing between
both phase centers in order to ensure maximum isolation
at minimum distance between the TX and the RX
antenna. The arrangement of the antennas and the RX
antenna pattern allow that GPS signals can be receipt
additionally to the pseudolite signals. Both TX antenna
and RX antenna operate at 1575.42MHz, since the low-
cost RXs in the system only support the reception of L1-
signals. Additionally to the navigation signal antennas,
there is a long-range WLAN antenna at each TC to cover
large operation areas. The antennas are mounted on
tripods at a height of about 2.50m above ground to
prevent signal shading by irregularities of the ground and
persons moving within the test area.
Fig. 1 Transceiver Hardware
The pseudolite signals are generated with the GNSS
signal generator NSG 5100 developed by EADS Astrium.
A detailed description of the NSG 5100 is available in
(Martin et al., 2007). The advantage of using the NSG
5100 as signal generator is that the system can easily be
switched from GPS pseudolite signals at L1 to Galileo
pseudolite signals at E5, E6 or E1. However, presently
the RXs used in the SekaN system only support the
reception of GPS L1 C/A code signals. As soon as
inexpensive Galileo RXs are available the SekaN system
can be operated at either the E5 or the E6 or the L1-band.
The NSG 5100 transmits pseudolite PRNs in order to
ensure that the SekaN system does not interfere with GPS
or other satellite-based navigation systems. Furthermore,
the NSG 5100 fits well for pseudolite applications
providing just one channel at a time. The precise 10MHz
clock (OCXO) of the NSG 5100 is also used as external
clock for the GPS/PSL Receiver whose internal clock can
be driven by a more precise external clock. The transmit
power of the NSG 5100 can be adjusted in a range from -
32dBm…0dBm. No additional navigation signal
amplifiers are required between the RF signal output and
the passive TX antenna.
All signal generators in the system are pulsing to
overcome the near-far problem in the local navigation
environment. The RTCM pulsing scheme as suggested in
(Stansell,1986) is applied. By emitting pulsed pseudolite
signals it is also ensured that satellite signals are not
jammed unintentionally.

Schlötzer et al.: Autonomous Navigation Environment with Self-Calibrating Transceivers 151
Master and Control Station
The SekaN system is controlled by the MCS which is
shown in Fig. 2. The main component of the MCS is a
high-performance laptop settled in a drawer of the 19”
MCS rack. A GPS/PSL receiver for surveying of the
pseudolite signals, a WLAN access point to communicate
with all SekaN system components and a power unit are
also integrated in the MCS rack. A long-range WLAN
antenna and a navigation signal RX antenna are mounted
on top of two separate tripods.
Fig. 2 MCS Hardware
The main tasks of the MCS are:
• System Configuration
• Self-Test (autonomous adjustment of the
pseudolite output power depending on the array
size and geometry)
• System Monitoring(surveying of house keeping
data, RX lock states and S/N)
• System Control (independent command
interface to the TCs and the ROV)
• Navigation Message Generation (provision of a
pseudolite navigation message to the signal
generators)
• System Self-Calibration
• Calculation of Rover RX Positions
• Visualization of TC and ROV Coordinates in a
Local Coordinate System or a Map
Fig. 3 shows the graphical user interface (GUI) of the
MCS:
Rover Receiver Architecture
The SekaN prototype system presently only comprises
one rover RX. The rover RX presented in Fig. 4 can be
separated into two modules: the rover vehicle module and
the user RX module.
Fig. 3 MCS Graphical User Interface
Fig. 4 Rover Receiver
The rover vehicle module can be replaced by any remote-
controlled vehicle which is robust enough to carry the RX
equipment. Alternatively, the user RX module can be
carried by a person, e.g. in a bag-pack. The user RX
module is comprised of a GPS/PSL RX, a navigation
signal RX antenna, a WLAN data link to the MCS and a
lightweight power unit.
The SekaN system does not make use of a rover
Transceiver to self-calibrate the system and to navigate in
the local navigation environment as suggested for the
Mars Navigation System presented in (LeMaster and
Rock, 2002). Using a RX-only module has the benefit
that the user module is inexpensive and that the
equipment can be miniaturized to a high degree.
Especially if many users are navigating in the
autonomous navigation environment, it is desirable to
keep the number of transmitters besides the stationary
TCs as low as possible to prevent signal interference.

152 Journal of Global Positioning Systems
4. REFERENCE SYSTEM
It is supported by the MCS to display the TC coordinates
and the user trajectory either in a local Cartesian
coordinate system or in a map which is referenced to
WGS84. Displaying of TC and user coordinates in an
absolute coordinate system (e.g. WGS84) is only possible
if there is any a priori information about the test area: If
the TC-array is planar, the coordinates of at least 3 TCs
have to be known in WGS84; else if the TC-array is
spatial, the coordinates of 4 TCs have to be known in
WGS84.
Referencing to an absolute coordinate system is also
possible if the TC-RXs in the system track GPS signals
besides the pseudolite signals. Single GPS solutions can
be derived from the satellite observations which may be
used for referencing the local coordinate system to
WGS84. However, this approach is not recommended
since the single GPS solution for the TC coordinates is
generally less precise than the self-calibration solution for
the local TC coordinates.
Fig. 5 Local Cartesian Coordinate System
Fig. 6 Referencing to WGS84
In Fig. 5 and Fig. 6 the self-calibrated TC-coordinates
and the rover RX trajectory (carrier phase solution) are
displayed both in the local coordinate system and in the
absolute coordinate system. Coordinate constraints are
introduced during the self-calibration process in order to
constrain the degrees of freedom of the network due to
translation and rotation. The self-calibration algorithm is
simplified by settling the TCs in the local coordinate
system as follows:
• TC1* is settled in the center of the coordinate
system (x1*=0, y1*=0, z1*=0)
• TC2* is settled on the positive x-axis of the
coordinate system (y2*=0, z2*=0)
• TC3* is settled in the xy-plane and has a positive
y-value (y3*>0, z3*=0)
If a three-dimensional TC-array has to be self-calibrated,
the following coordinate constraint is additionally
introduced:
• TC4* is settled out off the xy-plane and has a
positive z-value (z4*>0)
The TCs are marked with (*) to indicate that the TC-
indices used by the self-calibration are not necessarily
identical with the HW-indices of the TCs. If three TCs
are almost on a straight line as in case of TC1, TC2 and
TC3 in the current example (see Fig. 5), the algorithm
detects this automatically and chooses 3 other TCs to set
up the local coordinate system, e.g. TC3, TC4 and TC5,
which form a more regular triangle. In the current
example the algorithm resorts the TC indices as follows:
TC3 → TC1*, TC4 → TC2*, TC5 → TC3*, TC1 → TC4*,
TC2 → TC5*.
5. SELF-CALIBRATION ALGORITHM
The TCs in the SekaN navigation environment are self-
calibrated without any a priori knowledge about their
locations. Precondition for successful self-calibration of a
planar TC-array with the help of a moving user RX is the
availability of at least four stationary TCs. If the TC-array
is spatial, at least 5 stationary TCs are required for
successful self-calibration.
The schematic approach of the self-calibration algorithm
used for the SekaN system is presented in Fig. 7. The
coarse-calibration is based on observation data from a
static setup and only pseudorange measurements are
processed. The fine-calibration is based on observation
data collected by a rover RX during its trajectory inside
or outside of the TC-array. Only carrier phase
measurements are processed in the fine-calibration of the
system to derive the most precise solution. Unlike
previous investigations on self-calibrating TC-arrays
presented in (LeMaster and Rock, 2002) or (Matsuoka et
TC1
TC2
TC3 TC4
TC5

Schlötzer et al.: Autonomous Navigation Environment with Self-Calibrating Transceivers 153
al., 2002), this approach makes use of a rover RX-only
instead of a rover TC in order to collect the data required
for the fine-calibration of the TC-array. Thus, a new
approach has to be developed to derive the rover RX start
position and an estimate of the rover RX trajectory.
Afterwards, a nonlinear optimization is applied based on
double-differenced phase measurements collected during
the rover RX trajectory.
Fig. 7 TC-Array Self-Calibration Steps
The RXs in the system synchronize their sampling times
on a reference pseudolite PRN. After a software-based
fine synchronization of the RX raw data, the
measurements from different RXs can be differenced.
The clock errors cancel out in the self-differencing and
double-differencing process. It is therefore not necessary
to fine-synchronize the TCs in the network.
Self-differenced pseudorange observations between all
stationary TCs are used to calculate triangulation
solutions as first estimate for the TC coordinates:
)(2)( ,_, t
z
y
x
z
y
x
tji
i
i
i
j
j
j
ji ρ
νρ Δ∇+
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
−
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
⋅=Δ∇ (1)
, where:
ρ: pseudorange measurement
x,y,z: local Cartesian coordinates of the TCs
νρ: pseudorange noise due to all sources (e.g.
receiver noise, multipath)
The subscripts i and j are used to distinguish between two
transceivers TCi and TCj. Equation (1) has already been
simplified assuming that the TX antenna phase center and
the RX antenna phase center of the same TC are
collocated. As the clock errors cancel out in the self-
differencing process, the ranges between the stationary
TCs in the network becomes directly observable and
coarse estimates of the TC coordinates can be determined
via triangulation.
The rover RX start position is derived from a search in
the geometry domain. A two-dimensional or three-
dimensional mesh grid which covers the operation area is
set up. The mesh grid points represent candidates for the
rover RX start position under test. The sum of squared
measurement error residuals of the double-differenced
pseudoranges serves as test criterion.
An exemplary distribution of the sum of squared
measurement error residuals is shown in Fig. 8 (coarse
grid search) and in Fig. 9 (fine grid search) for a two-
dimensional test area. The fine grid search is performed
in a small subset of the original test area which is set up
around the selected test candidate of the coarse grid
search.
-100 -50 050 100 150
-100
0
100
200
0
1
2
3
4
x 10
4
X-Coordinate (m)
R ove r R ec eiver G r id Searc h
Y-Coordinate (m)
Sum of Squar ed Residuals
0.5
1
1.5
2
2.5
3
x 10
4
Fig. 8 Sum of Squared Measurement Error Residuals as a Function of
the Rover RX Position Estimate (coarse mesh grid)
45 50 55 60 65
-5
0
5
10
15
0
200
400
600
800
1000
1200
X-Coor dina t e ( m )
R o ver Rece iver Grid Se ar c h
Y-Coordinate (m)
Sum of Squ ar ed Res iduals
100
200
300
400
500
600
700
800
900
1000
Fig. 9 Sum of Squared Measurement Error Residuals as a Function of
the Rover RX Position Estimate (fine mesh grid)

154 Journal of Global Positioning Systems
The fine-calibration of the SekaN system relies on
double-differenced carrier phase observations which are
collected during the rover RX trajectory:
Fig. 10 Double-Differencing in the TC-Array
The double-differenced carrier phase measurements
φ
Δ∇ in units of meters can be formulated as follows:
)
()()( ,
,_
,
,
,
,
,
,tNtrt NREF
REF
NREF
REF
NREF
REF
NREF
REF
TCTC
TCRo
TCTC
TCRo
TCTC
TCRo
TCTC
TCRo φ
νλφ Δ∇+⋅Δ∇+Δ∇=Δ∇
(2)
with:
)()()(
,
,trrtrtr REFN
REF
NNREF
REF
TC
Ro
TC
TC
TC
Ro
TCTC
TCRo −−=Δ∇
REFN
REF
NNREF
REF
TC
Ro
TC
TC
TC
Ro
TCTC
TCRo NNNN −−=Δ∇ , ,
The assumption is made that the TX antenna and the RX
antenna of the reference TC are collocated. Thus, the
terms REF
REF
TC
TC
N and REF
REF
TC
TC
r cancel out in equation (2).
Furthermore, it is assumed that no cycle slips occur
during the rover RX trajectory so that the initial carrier
phase ambiguities stay constant. In contrast to satellite
navigation, the ranges between the reference RX and the
pseudolites are time-invariant. The unknown system
states, e.g. the TC coordinates, the rover RX coordinates
and the carrier phase ambiguities, are solved iteratively
during the nonlinear optimization process of the fine-
calibration. The initial state estimates for the nonlinear
optimization are derived from the previous self-
calibration steps. After a self-calibration solution has
been computed for the TC network, stochastic quality
checks are carried out. This is essential for an operational
system which shall provide a reliable navigation
environment.
6. USER POSITIONING ALGORITHM
The user positioning algorithm supports a stand-alone
positioning mode and a differential positioning mode.
The stand-alone positioning mode is presently only used
for GPS-only navigation as the TC network is only
coarse-synchronized. Working with unsynchronized
pseudolite signals implies that differential methods have
to be applied in order to cancel out the clock errors. The
RX of the reference TC, which is determined during self-
calibration, serves as reference RX in the differential
mode. Both satellite signals and pseudolite signals can be
processed by the user positioning module. However, in
order to combine both satellite signals and pseudolite
signals for the navigation solution, it is necessary to
reference the local coordinate system of the TC-array to
an absolute coordinate system. The self-calibration only
provides the TC coordinates in a local Cartesian
coordinate system. In order to perform a coordinate
transformation from the local coordinate system to an
absolute coordinate system, some of the TC coordinates
have to be available in the absolute coordinate system. As
there is generally no a priori information about the TC
locations, the user positioning module works with
pseudolite-only signals in the SekaN system. Then the
rover RX positions, the headings and the velocities which
are forwarded refer to the local Cartesian coordinate
system.
The flowchart presented in
Fig.11depicts the process of two-dimensional navigation
which has been investigated most intensely so far.
Fig.11 Flow-Chart of the User Positioning Algorithm
Provided that two synchronized data sets from a reference
RX and the rover RX are available, a differential code
solution is determined. A differential carrier phase
solution is additionally calculated if the float filter
converges (see Fig. 12).
Fig. 12 Flow-Chart of the Phase Solution

Schlötzer et al.: Autonomous Navigation Environment with Self-Calibrating Transceivers 155
7. SYSTEM TEST RESULTS
Test Area
For the system tests the SekaN TCs are set up in a
hexagon at plane grassland. The MCS is settled in the
center of the hexagon. However, the MCS can also be
placed outside of the hexagon as long as there is line of
sight to all SekaN system components. The side length of
the hexagon marked by the TCs is approximately 120m.
Fig. 13 TC Setup in the Test Field
In this test field only two-dimensional self-calibration and
user positioning is possible as the TCs and the user RX
are all almost in the same plane. The resulting VDOP
values in the operation area are too poor to make the z-
components sufficiently observable. If satellite signals
were tracked additionally to the pseudolite signals, also
three-dimensional user positioning would become
possible. Nevertheless, the self-calibration is still
restricted to two dimensions as no satellite signals are
used for the self-calibration process. The user positioning
can only work in pseudolite / satellite mixed mode if the
local TC network can be referenced to WGS84. This
implies that a priori information about some of the TC
positions is required which, in general, is not available.
In preceding simulations successful two-dimensional and
three-dimensional self-calibration has been demonstrated
for various TC constellations. Also the user positioning
algorithm has been tested successfully for two-
dimensional and three-dimensional applications.
However, in these system tests only pseudolite-only, two-
dimensional self-calibration and user positioning were
investigated. Therefore, the HDOP in the operation area
defined by the TCs (see Fig. 14) is decisive for the
quality of the position solutions.
-100
-50
0
50
100
150
200
-50
0
50
100
150
200
250
HDOP
x [m]
HDOP in the PSL Test Area
y [m]
1
1. 5
2
2. 5
3
3. 5
Fig. 14 HDOP in the Operation Area
Irregularities in the HDOP pattern presented in Fig. 14
result from different side lengths of the TC hexagon and
slightly different TX antenna heights. The HDOP in the
operation area varies between 0.87 in the center of the
hexagon and 3.36 in the direct vicinity of a TC. Generally
it is advisable to keep a minimum distance to the TCs to
prevent an enhancement of the position error. Pseudolite-
only navigation outside of the hexagon is also possible,
but the HDOP increases significantly if the rover RX
moves far away from the hexagon.
Self-Calibration Results
Fig. 15 shows the results of the self-calibration of a TC-
array consisting of 6 TCs. The self-calibration of the two-
dimensional system would already work with 4 TCs. It is
beneficiary to have redundant TCs in the system to
improve the performance of pseudolite-only user
positioning in the local navigation environment.

156 Journal of Global Positioning Systems
Fig. 15 Self-Calibration of 6 Transceivers
Green pluses mark the coarse-calibrated TC coordinates
in the local coordinate system of the MCS GUI. The fine-
calibrated TC coordinates are mapped as red pluses while
the actual TC coordinates are mapped as black pluses.
The fine-calibrated TC positions are almost collocated
with the actual TC positions which have been determined
geodetically a priori to the field tests in order to provide a
reference. A first estimate of the rover RX trajectory is
plotted as blue line in the coordinate system. The estimate
of the rover RX trajectory has almost the same shape as
the actual trajectory except for an offset of a few meters.
The rover RX trajectory used to fine-calibrate the system
is of arbitrary shape, however providing sufficient
relative range change between the stationary TCs and the
rover RX. The number of trajectory sample points is an
important factor for successful self-calibration of the TC-
array. On the one hand, the more sample points are
considered the more increases the computational time. On
the other hand, the less sample points are considered the
more likely it is that the self-calibration algorithm
diverges. For this concrete field test whose results are
presented in Fig. 15, the rover RX has sampled with 1Hz
during its trajectory. Altogether 329 trajectory sample
points were recorded during 5.5 minutes. All sample
points were considered for the fine-calibration of the TC-
array.
The time required for successful self-calibration did not
exceed 15 minutes in any of the field tests. Actually, the
TC-array could be self-calibrated within only 10.5
minutes most of the time. The composition of the self-
calibration times during field tests is indicated in
Table 1.
The accuracy of the coarse-calibration and fine-
calibration is presented in Fig. 16.
The maximum deviation between the coarse-calibrated
TC positions and the true TC positions is 9.62m for TC6*.
After the fine-calibration the deviations are smaller than
0.08m for all TCs in the network.
Table 1 Typical Self-Calibration Times
Self-Calibration Step Typical Time Span
Data Acquisition →
Coarse-Calibration
≈15s
Computing Time →
Coarse-Calibration
<1s
Data Acquisition →
Fine-Calibration
≈4.5min
Computing Time →
Estimate of the Rover Trajectory
≈20s
Computing Time →
Fine-Calibration
≈5.5min
∑ 10min 36s
123456
0
1
2
3
4
5
6
7
8
9
10
TC * Numb er
Range E rror [m]
Deviat ion from th e tr ue TC Loc at ions
c oarse c a l i bration
fine cali b rati o n
Fig. 16 Range Error between the coarse-calibrated (blue bars) / fine-
calibrated (red bars) TC Positions and the true TC Positions
Table 2 Deviation in x and y of the fine-calibrated TC Coordinates from
the true TC Coordinates
TC ID Δx [m] Δy [m]
TC1* 0.000 0.000
TC2* -0.028 0.000
TC3* 0.004 0.033
TC4* 0.030 0.058
TC5* -0.034 -0.072
TC6* -0.059 0.032
The accuracy of the self-calibration in this field test has
been better than 0.1m.
Several field tests have been carried out to derive
representative results: the number of TCs in the network,
the network geometry, the number of trajectory sample
points and the rover trajectory shape has been varied. As

Schlötzer et al.: Autonomous Navigation Environment with Self-Calibrating Transceivers 157
long as no cycle slips or signal shading occurred during
the rover RX trajectory, the accuracy of the self-
calibration was better than 0.3m. Altogether the
occurrence of cycle slips and signal shading was very rare
in this flat and obstacle-free test environment.
Furthermore, a cycle slip detection and correction
algorithm is implemented in the RX raw data decoders.
User Positioning Results
The evaluation of the user positioning performance is
more difficult than the evaluation of the self-calibration
performance. The rover RX antenna is not mounted
statically on tripods at known reference positions, but the
rover moves only coarsely to known reference marks.
There is an inherent uncertainty of several centimeters in
this approach when comparing the reference positions
with the calculated user positions. For the presentation of
the user positioning results in Fig. 17, the computed local
Cartesian rover coordinates are referenced to WGS84. A
map of the actual operation area can be uploaded in
which the computed user positions are displayed.
Fig. 17 User Positioning in the autonomous SekaN Navigation
Environment
The user RX moves in a triangle from U1 via U3 and U2
back to U1. The dark-blue doted trajectory indicates the
differential code-phase solution while the turquoise doted
trajectory indicates the differential carrier phase solution.
The accuracy of the user positions is better than 30cm
most of the time. If no differential phase solution can be
calculated, the accuracy of the user positioning gets
worse. In contrast to the differential phase solution the
differential code solution is rather noisy as low-cost
receivers are used with relatively high RX code noise
parameters. Presently no smoothing algorithm is
implemented for the differential code solution which
would reduce the noise of the code solution.
8. CONCLUSIONS
An operational autonomous navigation system has been
developed that covers new fields of applications. The
system comprises low-cost components and is extendible
to various numbers of TCs and users. Successful self-
calibration of TC networks consisting of 5 or 6 TCs
within less than 15 minutes has been demonstrated
several times during field tests. The accuracy of the self-
calibration is good enough to provide an autonomous
navigation environment where carrier phase based user
positioning becomes possible. The user positioning error
of the rover RX is lower than 30cm for pseudolite-only
differential phase solutions. Differential phase solutions
are available most of the time for the rover RX.
Further improvements of the SekaN system are possible
by making the self-calibration more robust to the
occurrence of cycle slips and signal shading if redundant
TCs are available in the system. Extensive field tests on
three-dimensional self-calibration and user positioning
are still required. However, good performance of the
system is also expected for three-dimensional
applications as various simulations have already yielded
in promising results.
ACKNOWLEDGMENTS
This research and the SekaN system development have
been conducted under DLR grants 50NA0503.
REFERENCES
LeMaster E. A., Rock S. M. (2002) An Improved Solution
Algorithm for Self-Calibrating Pseudolite Arrays.
Institute of Navigation National Technical Meeting, San
Diego, CA, January.
Martin S., Diefenbach S., Felbach D., Schlötzer S., v.
Voithenberg M. (2007) GNSS Signal Generator NSG
5100, Proceedings of the Institute of Navigation GNSS-
2007 Conference, Fort Worth, Texas, Sept.
Matsuoka M., LeMaster E. A., Rock S. M. (2002) 3-D
Capabilities for GPS Transceiver Arrays. Proceedings of
the Institute of Navigation GPS-2002 Conference,
Portland, OR, Sept., pp. 824-834.
Stansell T. A. (1986) RTCM SC-104 Recommended Pseudolite
Signal Specification. RTCM-/SC104-Std, Torrance, CA,
June.