2012 12th International Conference on ITS Telecommunications

Performance Evaluation of Real-time MEMS INS/GPS Integration with

ZUPT/ZIHR/NHC for Land Navigation
IJ
Cheng-Yueh Liu , Cheng-An Lin1, Kai-Wei Chiang1, Shih-Ching Huang2, Chin-Chia Chang2, Jia-Ming Cai2
IJ
Department of Geomatics, National Cheng Kung University, Taiwan
ij
Industrial Technology Research Institute, Taiwan

ABSTRACT GPS and INS. GPS can provide continuous, accurate

INS/GPS integrated system has been one of the most
positioning with lines of sight to more than four satellites.
popular methodologies in the research field of navigation
However, the accuracy and availability of GPS-based
technology. This paper aims at developing a real-time
vehicular navigation systems are subjected to the
low-cost INS/GPS integrated navigator based on PC
open-sky condition and degrade in the presence of signal
platform and reviewing many problems encountered in
blockage and reflected signals. INS can fill the GPS gaps
the development of real time system. A MEMS IMU is
to provide continuous navigation solutions. In principle,
applied in the system to meet the requirements of low
IMU refers to inertial sensors including three gyroscopes
cost and small size. The proposed system utilizes Auto
and accelerometers and it provides compensated raw
ZUPT/ZIHR as an accurate update source for EKF to
measurements including velocities changes (delta-Vs) and
improve the accuracy of navigation solutions during GPS
orientation changes (delta-ȧs) along three directions of
signal blockages. In the end, various field tests which
its body frame. Although an integrated navigation system
include a pure INS aided by ZUPT/ZIHR and NHC, a
can work in GPS-denied areas, problems including the
simple INS/GPS and an INS/GPS aided by ZUPT/ZIHR
cost of inertial sensors and the length of time during the
and NHC are conducted in downtown. The preliminary
unavailability of GPS signals affect its applicability. For
results illustrates the proposed system with frequent
this reason, MEMS inertial sensors are preferred as the
ZUPT/ZIHR provides a stable performance operating
complementary component to GPS for seamless vehicular
kinematically in free inertial mode without the aiding of
navigation applications.
GPS by 15 minutes with a low-cost IMU and the
Real-time INS/GPS integrated navigation system has
INS/GPS integrated solutions are improved by 50% when
been investigated with various computation platforms (Li,
automatic ZUPT/ZIHR and NHC are applied in GPS
et al., 2008; Zhang, 2009). In this study, PC is chosen as
denied areas.
the kernel platform to implement the real-time INS/GPS
Keywords: Real-time, INS, GPS, ZUPT, ZIHR, NHC
integrated navigation system, but there are still some
INTRODUCTION challenges. Therefore, the first objective of this study is to
The last two decades have shown an increasing trend in perform system integration in hardware and provide
the use of positioning and navigation technologies in effective solutions dealing with measurement
vehicular applications. Because of the importance of synchronization. Several INS/GPS integration schemes
location information, the market has pushed hard for the are available for implementing seamless land vehicular
development of reliable vehicle navigation and guidance navigation applications, such as loosely coupled and
systems, providing location information. Typical tightly coupled integration schemes (Titterton and Weston,
candidates for integrated navigation system to provide the 2004). Although tightly coupled has advantages including
position, velocity and heading in a seamless fashion are high accuracy and the capacity of updating GPS when the

978-1-4673-3070-1/12/$31.00 ©2012 IEEE 584

number of satellites is less than four, the complexity of applied for DGPS processing with single frequency carrier
computation produces a heavy computation load for phase measurements providing positioning accuracy for
processors (Niu and Han, 2008). High fault tolerance and post processing applications, as shown in Figure 2
low numerical computation burdens make loosely coupled
Unified to RS232
easier to implement (Titterton and Weston, 2004). Communication
Loosely coupled is thus chosen here due to its simplicity Protocol
PC
of implementation and good performance. The
IMU SPI to RS232 ( 32-bit,
disadvantage of implementation is that measurement
(done by MCU) 2.0 GHz,
update of integrated navigation system is only possible
2 GB
when four or more satellites are in view. Therefore, the RAM)
second objective of this study is to modify a conventional GPS USB to RS232
loosely coupled INS/GPS integration scheme and
implement various adding functions including auto Zero Figure 1. The Integration block diagram and picture of proposed

Velocity Update (ZUPT), Zero Integrated Heading Rate INS/GPS integrated system

(ZIHR) and Non-Holonomic Constraint (NHC) as the

primary navigator for the real-time INS/GPS integrated
navigation system to improve the positioning and
orientation accuracy during long GPS outage. Ultimately,
the third objective is to verify the performance of
proposed system by conducting in downtown areas. The Figure 2. The chosen IMU from ADI 16385 and GPS module from

analysis is conducted concerning the impact of aiding ublox

from ZUPT/ZIHR and NHC, respectively.
THE CONFIGURATION OF PROPOSED INS/GPS
INTEGRATION SCHEME
! Hardware and Software Configuration
Three components including IMU, GPS and PC are
considered as the core of the hardware architecture of
proposed INS/GPS integrated navigation system, which is
shown in Figure 1. The proposed low-cost PC-based
system mainly involves a commercial laptop to integrate
a low-cost MEMS tactical grade IMU which outputs
1KHz digital signals of acceleration and angular rate and
one commercial 10Hz GPS module. The sensors are
On the other hand, there are three primary challenges
concerning the implementation of real-time INS/GPS
integrated navigator. The first one is strict restriction of
execution time for mathematical computations including
INS mechanization equations and sensor fusion engine;
the second one is concurrency control of shared resource
when utilizing multi-thread programming model to
realize multiple senor data acquisition with various
sampling rates. The third one is time synchronization for
multiple sensor data.
! Kernel Computation module
The kernel computation module accounts for the

dynamically compensated and provide a simple user
interface and control via a standard SPI (serial peripheral
interface) and register interface, and the specifications of
GPS receiver, ublox AEK-6T. This module is chosen
overall mathematical computations including strap-down
INS mechanization and EKF with a closed loop loosely
coupled INS/GPS integration.
! Multi-thread Programming for Multiple Sensor

because it can provide L1 carrier phase raw measurements Data Acquisition

585
 To implement data acquisition module for different Figure 4. Software-based time synchronization scheme
sampling rates sensors, each module in the system is
implemented in a single thread. Multi-thread model of
proposed real-time INS/GPS integrated navigator is
illustrated graphically in Figure 3. The shared data can be
well managed with the concurrency control functions.

OS on PC platform

Thread for Thread for

GPS Data IMU Data
acquisition acquisition
module module Figure 5. Hardware-based time synchronization scheme

Lagrange Interpolation is applied which is shown in

shared data : Eq(1). After interpolating INS solution to the desired time
sensor
point, there is difference between state at GPS update
epoch and state at previous epoch. When fusing the data,
Thread for Kernel computation
the difference must plus onto current state and replace it.
Figure 3. The multi-thread model
*+, *+,
t "#$ ( t )
! Time Synchronization of Sensor Data ! (t
"#$ ) % !
r (t & ) '
t& ( t)
r =
In integrated system, all time tags should reference to &-*.,./ )-*.,./
)0&
identical time system for further processing. Time
3456 .37
v ! (t "#$ ) = 1&-*.,./
*+,
v ! (t & ) 2)-*.,./
*+,
(1)
synchronization problem can be classified into two 38 .37
)0&
categories. One indicates the bias and scale factor of time
where r ! is position vector and v ! is velocity vector of
system of device is different from each other. The other is
INS in navigation frame, t & is i-th time epoch in system
misalignment of sampling timing between sensors. Thus,
time frame, and t "#$ is time epoch of updating by GPS.
GPS and IMU can fuse navigation solution nearly at the
! Proposed INS/GPS Integration Scheme
same epoch so mathematical model of updating function
The information flow of integrated navigation system
can be properly applied. Figure 4 shows software-based
with an EKF is shown in Figure 6, and it estimates true
method, and Figure 5 is illustration for hardware-based
state directly and not the deviation of nominal trajectory.
time synchronization scheme of IMU measurements.
The estimated state is fed back into INS to process
Time tagging mechanism is maintained by PC timer
nominal trajectory ahead which is recursively estimated
program to tag all measurements provided by IMU and
by EKF. Figure 7 illustrates proposed closed loop loosely
GPS. When EKF utilizes GPS solutions to update INS
coupled INS/GPS integration scheme.
states, navigation solution of INS is interpolated into
INS position, Corrected
velocity, inertial
sampling time point so updating effect is validly applied. IMU Acceleratio INS etc. + errors output
Angular Velocity
GPS
Estimated
position,
velocity,
INS INS INS INS INS INS attitude,
errors

Position + Kalman
GPS receiver Filter
t *.: t *.9 t *./ t* t *+/ t *+9

Figure 6. Information flow with EKF

586
 Gravity ZUPT/ZIHR mode automatically. It raises challenges to
Accelerometers
(ː v)
Corrections ! ! Positions judge whether to trigger ZUPT/ZIHR or not.
RESULTS AND DISCUSSIONS
Velocities

Gyroscopes
To verify performance of proposed system, kinematic
(ː !"
Corrections ! Attitudes
field tests were conducted around urban canyon in
Auto NHC INS EKF downtown. The performance evaluation is conducted by
Auto ZUPT
Auto ZIHR
real time navigation solutions provided by INS/GPS
GPS EKF
GPS (position/velocity) integrated navigator and compare with the reference
Figure 7. Proposed loosely coupled INS/GPS integration scheme trajectory processed with RTS smoother in post-mission.
Auto Non-Holonomic Constraints Experiment were used for 3 different scenarios including

The velocity of land vehicle navigation constraints are
derived assuming that vehicle does not slip and stays on
the ground. If both assumptions are true, NHCs are
defined, velocity of vehicle in the plane perpendicular to
forward direction is almost zero (Sukkarieh, 2000).
Figure 8 shows two NHCs in the b-frame considered as
kinematic test for pure INS aided by arbitrary-
ZUPT/ZIHR and NHC (Test-1), INS/GPS integrated
system (Test-2) and INS/GPS integrated system aided by
arbitrary-ZUPT/ZIHR (Test-3). Because test van cannot
stop too frequently, the benefit of ZUPT/ZIHR was not
significant. When NHC is applied, accuracy of proposed
system is improved by 50%. Trajectories of tests can be

measurement updates to EKF. seen in Figure 9, and error variation processes of position

<=> ? 0A
and attitude are shown in Figures 10, 11, 12. Statistical
;
<@> ? 0 analysis results are shown in Tables 1.

Figure 8. Two non-holonomic constraints in the b-frame

Auto ZUPT/ZIHR

In inertial mode, ZUPT means the occasion stop of

system for short duration estimating errors of system and
bounding the growth of inertial sensor errors. If vehicle
stops, velocity outputs in any directions should be zero.
Figure 9. The kinematic trajectory of tests
On the other hand, ZIHR utilizes the fact that if heading
angle computed by INS can be updated by the first-stored
heading angle which remains as instantaneous heading
angle when vehicle stops. Whenever ZUPT initiates,
ZIHR follows. In this study, auto ZUPT and ZIHR are
triggered by a threshold by observing accelerometers
outputs through window based approach automatically.
When vehicle is not leveling, gravity vector will project
onto three axes, which prevents system from entering
Figure 10. Error variation processes of position and attitude of Test-1

587
 CONCLUSIONS
This study exploits several critical issues concerning
implementation of real time INS/GPS integrated
navigator. A MEMS IMU is applied in system to meet the
requirements of low cost and small size. The proposed
system utilizes auto ZUPT/ZIHR as an accurate update
source for EKF to improve accuracy of navigation
solutions during GPS signal blockages.
Figure 12. Error variation processes of position and attitude of Test-2 Preliminary results illustrates the proposed real time
system with frequent ZUPT/ZIHR provides a stable
performance which operates kinematically in free inertial
mode without GPS by 15 minutes with a low cost IMU.
On the other hand, real time INS/GPS integrated
solutions are improved by 50% when auto ZUPT/ZIHR
and NHC are both applied in GPS denied areas.
REFERENCES

! Titterton DH, Weston JL (2004): Strapdown inertial navigation

Figure 12. Error variation processes of position and attitude of Test-3 technology, 2nd edition, American Institute of Aeronautics and

Table 1. Statistical error analysis Astronautics, Reston, USA.

RMS of Position Error (m) ! Niu X, Sameh N, Goodall C, El-Sheimy N (2007): A universal

North East Height approach for processing any MEMS inertial sensor configuration

Test-1 70.1122 195.6053 21.3215 for land-vehicle navigation, Journal of Navigation, vol. 60, issue 02.

Test-2 46.7511 28.1232 15.5441 ! Chiang KW, Noureldin A, El-Sheimy N (2003) : Multisensor

Test-3 46.3741 15.6062 6.3152 integration using neuron computing for land-vehicle navigation.

GPS Solution 6(4):209-218. DOI:10.1007/s10291-002-0024-4.

RMS of Velocity Error (m/s)
! Goodall C (2009): Improving usability of low-cost INS/GPS
North East Height
navigation systems using intelligent techniques. PhD thesis,
Test-1 1.0912 1.2474 0.1117
University of Calgary.
Test-2 2.1406 1.4911 0.5311
! Silberschatz Abraham, Galvin Baer Peter, Gagne Greg (2009):
Test-3 2.1445 0.8290 0.1310
Operating System Concepts, 8/e., Wiley.
RMS of Attitude Error (deg)
! Conte, SD and de Boor, C (1980): Elementary Numerical Analysis:
Test-1 0.3610 0.8237 5.7521
An Algorithmic Approach. McGraw–Hill, third edition.
Test-2 0.8939 1.2909 3.4302
! Shin Eun-Hwan (2005): Estimation Techniques for Low-Cost
Test-3 0.8833 1.3350 4.8904
Inertial Navigation. UCGE Reports Number 20156.
Generally speaking, the proposed real time INS/GPS
! Grewal, Mohinder S., Andrews Angus P. (2008) : Kalman Filtering
integrated navigator aided by auto ZUPT/ZIHR and NHC
Theory and Practice Using Matlab, John Wiley & Sons Inc.
can provide 40% and 60% improvement in terms of
! Li Yong, Mumford Peter and Rizos Chris (2008): Performance of a
horizontal and vertical errors, respectively compared to
low-cost field re-configurable real-time GPS/INS integrated system
conventional.
in urban navigation, IEEE/ION 2008.

588