See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4269038

DGPS, RTK-GPS and StarFire DGPS performance under tree shading

environments

Conference Paper · April 2007

DOI: 10.1109/ICITECHNOLOGY.2007.4290370 · Source: IEEE Xplore

CITATIONS READS
40 3,441

2 authors, including:

Y. Morales
Standard Cognition
99 PUBLICATIONS   1,610 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Interaction between pedestrians and the automated vehicle (Human centered) View project

Artificial Consciousness Project View project

All content following this page was uploaded by Y. Morales on 05 February 2014.

The user has requested enhancement of the downloaded file.

 DGPS, RTK-GPS and StarFire DGPS Performance
Under Tree Shading Environments
Yoichi Morales and Takashi Tsubouchi
Graduate School of Systems and Information Engineering
University of Tsukuba, Tsukuba, 305-8573, Japan
Email: {yoichi,tsubo}@roboken.esys.tsukuba.ac.jp

proach which is a reliable method if a vehicle

Abstract— In this paper we present our experimental mea-
surement results using GPS in outdoor tall obstacle free and
can achieve self localization on the map. For this
under tree shading environments. We tested and compared
reason a reliable and robust localization method
three different GPS receivers using different principles such as
is an essential element for map based navigation.
Differential GPS, Real Time Kinematic GPS and an innovative
system called StarFire DGPS. Coordinate conversion issue is
As GPS is one of the most common approaches
discussed and stationary and moving tests were performed. In the
for outdoor localization on this paper, we report
stationary tests, off-line post-processed data was used to compare
the difference between receivers measuring the same point at
its performance under tree foliage environments.
different times. Finally three GPS receivers were mounted on a
Previous work has been done at university campus
mobile platform to perform moving experiments simultaneously
between buildings as addressed by Ohno et al.,
at open field and under tree foliage where experimental measure-
ments show the performance of each GPS receiver according to
its measurement principle in harsh conditions. ([1] and [2]). However, in the research of this
Keywords: GPS, outdoor, localization, StarFire, paper we concentrate on woodland environments.
tree canopy We consider that the first step is to test sensor per-
formance through tree shadow environments which
present not optimal conditions. Concretely we test
performance of common GPS technologies such as
I. I NTRODUCTION DGPS, RTK-GPS and an innovative high precision
Accurate outdoor localization keeps to be one of global DGPS variation such as StarFire.
the major tasks in many fields such as mapping, In section II, a brief explanation about GPS is pre-
navigation, machine automation and mobile robots. sented mentioning basic concepts and its drawbacks,
To achieve localization, self-contained methods such also methods to improve accuracy such as DGPS,
as odometry and inertial navigation using inertial RTK-GPS and a new technology called StarFire.
measurement units are commonly used, however as Coordinate conversion for GPS use is treated in
it is well known, their accuracy decrease with time section III. Section IV presents GPS receiver char-
as travel distance increases. The use of GPS as an acteristics. Section V presents experimental setup
external option for outdoor localization is a popular and experimental results. Conclusions are presented
issue. Availability of technologies such as DGPS, in section 6.
RTK-GPS and new technology such as StarFire
made possible localization on outdoor open field II. GPS: G LOBAL P OSITIONING S YSTEM
environments within centimeter order. GPS offers
global positioning and does not present accumula- The Global Positioning System is a satellite based
tive errors which is an attractive solution to correct positioning system developed by the Department
accumulated errors from dead reckoning methods. of Defense of the United States constituted of a
constellation of 24 satellites with an orbit around
A. Research Motivation the world of 12 hours and their ground stations. It
Our research motivation is the automation of can provide users with 3D position, velocity and
real construction machines in woodland mountain time. Common standalone GPS uses C/A code only
environments. In order to achieve autonomous nav- for position calculation, and has an accuracy within
igation, we plan to implement a map based ap- 10 meters. It is used for applications that only
need an approximate position as for example in car It uses C/A code and carrier phase for position
navigation systems. calculation. RTK needs initialization time of about
Data received by user from GPS receivers are 1 minute in order to give maximum precision. RTK
raw measurements and NMEA sentences [3] which offers two types of solutions, float and fix. RTK
contain information such as position, time, GPS float solution needs at least 4 common satellites and
fix data, satellite information, dilution of precision offers an accuracy range of about 20cm to 1m. RTK
(DOP) values, etc. There are several kinds of NMEA fixed solution needs at least 5 common satellites and
sentences. For example, they begin from GGA, GST offers accuracy within 2cm.
or GSA for positioning and receiver noise status. D. StarFire Differential GPS System
A. GPS Drawbacks StarFire is a global DGPS system developed
As any sensor, GPS has also its drawbacks and by NavCom Technology, Inc and Ag Management
accuracy and availability depend on many factors Solutions, it provides sub-decimeter high accuracy
(e.g. [4] and [5]): horizontal positioning [7], [8] and [9]. In order
• Tall obstacles around GPS antenna causing sig-
to obtain high accuracy, StarFire is based on a
nal block or multipath (signal reflection because technology called RTK (Real Time GYPSY) de-
of tall obstacles near user receiver) veloped by the Jet Propulsion Laboratory (JPL) for
• Weather conditions
the National Aeronautics and Space Administration
• Availability of at least four satellites in the
(NASA). The system is mainly constituted of seven
antenna view range for 3D localization. components: reference network, processing hubs,
• Satellite ephemeris
communication links, land earth stations, geosta-
tionary satellites, monitors and dual frequency user
There are many methods in order to improve receivers.
GPS accuracy, in next subsections three of them are
briefly explained. III. C OORDINATE C ONVERSION
GPS position information is given in latitude,
B. Differential GPS
longitude and ellipsoidal height, information in the
Differential GPS was created to correct bias errors WGS84 (World Geodetic System defined in 1984).
of the user receiver using measured bias errors at a For user localization applications, ellipsoidal lat-
known position. A base station computes corrections itude and longitude are difficult to interpret, so
for each satellite signal and sends it to the user. coordinate conversion is done in order to have
DGPS is mainly composed of 3 elements, one GPS position on an x-y Cartesian plane on the surface
receiver(antenna) at a known location (named base of the world ellipsoid. A planar Cartesian plane
station), one GPS receiver at an unknown location changes from region to region where the task is
(user receiver) and a communication medium be- to define the ellipsoidal world by a large amount
tween this two receivers. Base Station and user of small planar planes. For example, in the case of
receiver have to be within a distance of 100 km Japan, each prefecture enters into one of nineteen
and it needs at least 4 common satellites in view. planar Cartesian planes. Japanese Geographical Sur-
Accuracy of modern DGPS receivers is within 1 vey Institute (JGSI) defines the origin of each plane
meter. It uses C/A code for positioning. DGPS and provides coordinate conversion information [6].
can correct error sources with the exception of The originin of our coordinate system is situdated
multipath. at 139◦ 50’ 0” E in Longitude and 36◦ 0’ 0” N in
C. RTK-GPS
Latitude which is in calle as coodrinate system IX
of the JGSI .
Real Time Kinematic GPS can provide centimeter
accuracy measurements in real time. User antenna IV. GPS R ECEIVERS
needs to be within a distance of 10km to the We made measurements using 3 different re-
base station for receiving real time radio links for ceivers such as Trimble’s DSM12 DGPS, Trimble’s
information transmission and position correction. 5700 RTK-GPS and NavCom’s SF-2050M StarFire
DGPS. Each receiver has its own characteristics and A. Stationary measurement test in outdoor obstacle free envi-
calculates position using different principles. ronment
A. Trimble DSM12 DGPS Stationary measurements were done in outdoor
DSM 12/212 DGPS receiver has 12 L1 C/A code obstacle free environment (parking lot) over one
carrier phase channels, it includes and integrated hour period. Readings were taken at different times
dual-channel low noise MSK beacon receiver for measuring the same point. Measurement results are
differential corrections. The time to first fix is within shown in table 1 where standard deviations show
30 seconds. Differential position accuracy according the precision of each measurement at the 95%
to catalog specs is less than 1 meter horizontal RMS. confidence level. Figure 1 shows GPS plotted data
Antenna type: Dome. in x-y coordinate system (1m x 1m). It has to
B. Trimble 5700 RTK-GPS
be mentioned that the present authors have made
several measurements of the same point at different
Trimble 5700 RTK GPS receiver has 24 dual- days and times obtaining similar results confirm-
frequency channels (L1 C/A Code, L1/L2 Full Cy- ing the repeatability of stationary measurements.
cle Carrier, WAAS EGNOS). According to catalog Measurements were done on November 21st and
specs it has an horizontal position accuracy of ± 22nd of 2006 during clear sky days. Post-processed
1cm RMS and a vertical position accuracy of ± data was calculated using raw data from NavComs’s
2cm RMS. Antenna type: Zephyr Geodetic receiver with SOPAC’s (Scripps Orbit and Perma-
C. NavCom SF-2050M nent Array Center) SCOUT interface which is an
SF-2050M GPS receiver has 26 tracking channels ITRF coordinate Generator [13] projected on the
(12 L1/L2 full wavelength carrier phase tracking ellipsoid of WGS84. Post-processing is calculated
GPS + 2 dedicated SBAS), C/A P1 and P2 code offline taking several minutes. We consider this
tracking. It has a tri-band antenna which can receive post-processed data as the most accurate one and
GPS and StarFire signals. It is specified in technical use it as our ground truth value. DGPS, RTK-GPS
specs that it has an accuracy using real-time StarFire and StarFire data were calculated using the on-
system of within 10cm in horizontal position and line algorithms of the receivers using GGA NMEA
within 15cm (rms) in vertical position. Antenna sentence and coordinate conversion explained on
Type: Triband Dipole. section III. DGPS and RTK-GPS data was taken
V. E XPERIMENTAL S ETUP AND R ESULTS
after twenty minute warming up and StarFire log
was taken after initialization time of one hour.
These GPS receivers were tested stationally in an Logging rate was set to 1 Hz.
open field and moving in open field and under tree
foliage as explained in next subsections.Before ex- Used Method
SOPAC Trimble Trimble5700 NavCom
perimental explanation, we define some parameters SCOUT DSM12/12 RTK-GPS SF-2050M
to measure GPS receiver performance: Parameter Post - DGPS fix RTK fix StarFire RTG
Processed dual freq
• Accuracy refers to the closeness of sample mean
# of Sat 9 to 10 9 to 10 9 to 10 9 to 10
to the true value HDOP —– 0.90 to 1.2 1.1 to 1.2 0.90 to 1.5
x average [m] 23885.71 23885.88 23885.77 23885.76
• Precision refers to the closeness of the observa-
y average [m] 12465.09 12465.07 12465.16 12464.97
tions to the observation sample mean z average [m] 69.07 69.32 69.05 69.17
2σx [m] 0.0306 0.2448 0.0080 0.0236
• Continuity is the precentage of time that a
2σy [m] 0.0514 0.4304 0.0114 0.0398
system is performing a required function under 2σz [m] 0.1120 0.8594 0.0318 0.0636
stated conditions [10] and [11]. TABLE I
• Reliability is the probability that a service, when GPS STATIONARY MEASUREMENT RESULTS : N O OF SATELLITES , HDOP,
it is available, performs a specified function AVERAGE AND 2σ VALUES (95% CONFIDENCE )

without failure under given conditions for a

specified period of time. [12]
For stationary measurements we measured precision Difference between post-processed average point
and for moving tests we measured continuity. and GPS receivers on-line measured average points
 Fig. 1. Stationary Measurement Results

were of 17.11cm, 9.21cm and 13cm for DGPS, Fig. 2. Three different receivers and their antennas on Mobile Robot. From
RTK-GPS and StarFire respectively. We must be left to right RTK-GPS, StarFire, DGPS
aware of these observed offsets when we use GPS
even if we use RTK-GPS which provides the nar-
rowest deviation so there is the case that measured speed. Results of 10 experimental measurements are
data contains such offset. presented on table II showing position calculation
continuity, the number of satellites and HDOP val-
B. Moving measurements in outdoor tree foliage environments ues of each section during moving tests.
For moving measurements, GPS receivers were 1) DGPS Measurement: DGPS receiver robustly of-
placed on a fixed spot on a mobile robot platform, fered data under tree foliage, however, as it can
one next to each other (Figure 2). Measurements be seen in Figure 4, there are some data jumps
were done on November 3rd 2006 from 11:00 hours through section B-C (in the order of meters). Bea-
in a clear sky day at Tsukuba University’s parking con receivers work at LF frequencies (283.5 to
lot next to building L and woodland path near a 325KHz), for this reason even in cluttered environ-
parking Lot. ments, waves can wrap around objects and arrive to
For moving tests, start position was selected to be their destination. DGPS can initializate immediately
an obstacle free environment, i.e., after receivers are making position data available if there are enough
initialized and receiving fix measurements mobile satellites for positioning
platform was first manually pushed starting from 2) RTK-GPS Measurement: RTK-GPS offers very
point A to point B in Figure 3, then under tree precise measurements with no tall obstacles around
canopy path from point B to C finally ending on (Figure 5), however as robot entered tree foliage,
open field returning to start point A. Experimental mode quickly changed from fixed to float solution.
path is shown on Figure 3 which is a satellite picture This mode requires 5 satellites to initialize, taking
taken from Google earth API [14] (trajectory was about 60 seconds to initialize or to regain fix mode.
drawn by hand). Measured traversed trajectories can These are the reasons why RTK-GPS use under tree
be seen on Figures 4, 5 and 6. Under tree foliage foliage is complicated.
section, points were carefully measured and plotted 3) StarFire Measurement: StarFire provides two
to show real path of robot trajectory. Total path kinds of measurements, single frequency (with pre-
was approximately of 305 meters traveled at 1m/sec cision within 2m) and double frequency measure-
Fig. 3. Path in bird view. Parking lot with tree canopy passage
(http://www.roboken.esys.tsukuba.ac.jp/ yoichi/Map)
Fig. 5. Moving RTK measurements

Fig. 4. Moving DGPS measurements

Fig. 6. Moving StarFire measurements

ments (centimeter precision). After single frequency dual freqency) covariance ellipses are barely visible.
is aquired, it takes approximately 100 seconds to When it entered tree area, there are data jumps and
achieve double frequency status. At open field envi- there is non fixed data with small covariances and
ronments StarFire operated in dual frequency mode, fixed data with large covariance ellipses. As soon
however, when it entered section B-C because of as differential correction information was received
GPS and Geostationary satellite signal blockage outside woodland path, StarFire fixed solution was
StarFire DGPS fix data quickly dropped to no GPS achieved in single frequency mode.
fix mode where position could not be continously
computed as can be seen on Figure 6. An illus- VI. C ONCLUSIONS
tration showing StarFire measurements with their In this paper we presented an overview of GPS
respective covariance ellipses calculated using GST system, we mentioned its characteristics and some
NMEA sentence is shown on Figure 7 where it of its drawbacks, moreover, DGPS, RTK-GPS and
can be seen how before entering tree area, (in StarFire systems were briefly described. In addi-
 stacles nearby, because of cycle slips, fixed solution
drops to float solution producing some data jumps
as well. When there are no ideal conditions RTK-
GPS needs initialization to achieve fixed measure-
ments again, which is not appropriate for real-time
applications.
New technology such as StarFire in open sky
provided an accuracy and precision close to RTK-
GPS. As StarFire receives correction signals using
geostationary satellites (L-band), when signals are
blocked by tall obstacles, it does not provide good
continuity solution. StarFire advantage is that it can
be used anywhere without considering local base
station on earth as it receives correction information
from Geostationary satellite.
ACKNOWLEDGMENTS
This work has been supported by the Japan Soci-
Fig. 7. StarFire under tree foliage with covariance ellipses ety of Promotion of Science (JSPS) Grant-in-Aid for
Scientific Research (Scientific Research (B)) under
GPS Receiver
contact numbers 18360116. Authors would like to
Trimble Trimble5700 NavCom thank Mr. Masayuki Uchida and Mrs. Elaine Yuen
DSM12/12 RTK-GPS SF-2050M
Continuity % (A-B,C-A) 100 % 100% 100%
from GNSS Technologies Inc. for their technical
Continuity % (B-C) 100% 65.15% 54.76% support on GPS.
# of used Sat (A-B,C-A) 8 to 9 6 to 7 8 to 9
# of used Sat (B-C) 6 to 5 3 to 4 3 to 5 R EFERENCES
HDOP (A-B,C-A) .90 to 1.0 1.0 to 1.2 1.2 to 1.6
HDOP (B-C) 1.2 to 2.4 4.4 to 7.5 5.0 to 15.4 [1] K. Ohno, T. Tsubouchi, B. Shigematsu and S. Yuta “Outdoor Navigation
of a Mobile Robot between Buildings based on DGPS and Odometry data
TABLE II Fusion”, Proc. of IEEE Int Conf. on Robotics and Automation (2003)
R ESULTS OF 10 MOVING MEASUREMENT TESTS 1978-1984
[2] K. Ohno, T. Tsubouchi, B. Shigematsu and S. Yuta. “Mobile Robot
Navigation Based on DGPS and Odometry in Campus Environment”,
Proc. of IEEE Int Conf. on Field and Service Robotics (2003) 189-194
[3] http://www.nmea.org/pub/0183/
[4] K. Ohno, T. Tsubouchi, B. Shigematsu and S. Yuta. “Proper Use of GPS
tion, coordinate conversion issue was explained. for Outdoor Navigation by an Autonomous Mobile Robot”, Proc. of Int.
Conf. on Intelligent Autonomous Systems (2004) 1070-1079
Stationary measurement results showed how in [5] J. Huang and H. S. Tan. “A Low order DGPS-Based Vehicle Position-
open sky RTK-GPS offered the most precise so- ing System Under Urban Environment”, IEEE/ASME Transactions on
Mechatronics, Vol.11. No.5, October 2006
lution followed by StarFire and DGPS respectively. [6] http://vldb.gsi.go.jp/sokuchi/
Measurements of section 4A showed a small but [7] R. R. Hatch and R. T. Sharpe “Recent Improvements to the StarFire
Global DGPS Navigation Software” Journal of Global Positioning Sys-
clear difference (offset) between measurements of tems (2004) Vol.3, No.1-2:143-153
three receivers and post-processed calculation even [8] K. Dixon “StarFire: A Global SBAS for Sub-Decimeter Precise Point
Positioning”, Proceedings of ION GNSS 2006
though the same point was carefully measured. [9] R. Hatch, T. Sharpe and P. Galyean. ”StarFire: A Global
High Accuracy Differential GPS System”, Technical paper.
When high precision and accuracy measurements http://www.navcomtech.com/archives.cfm
are required using different receivers, GPS users [10] T. Moore, C. Hill and L. S. Monteiro “Maritime DGPS: Ensuring the
Best Availability and Continuity” THE JOURNAL OF NAVIGATION
have to be deeply concerned about this issue. (2002), 55, 485-494.
On moving measurements under tree canopy, [11] IMO (2001). Revised Maritime Policy for a Future Global Navigation
Satellite System (GNSS). Resolution A.915(22), 29 November 2001.
DGPS despite being less accurate and precise, [12] IMO (1997). Maritime Policy for a Future Global Navigation Satellite
proved to be the most robust solution offering System (GNSS). Resolution A.860(20), 27 November 1997.
[13] http://sopac.ucsd.edu/
position availability data during larger periods of [14] http://www.roboken.esys.tsukuba.ac.jp/ yoichi/Map
[15] G. Strang and K. Borre “Linear Algebra, Geodesy, and Gps”, Wellesley-
time. Cambridge Press 1997
In the case of RTK-GPS when there are tall ob-

View publication stats