Winning Student Paper - The IEEE Oceanic Engineering Society

Winning Student Paper - The IEEE Oceanic Engineering Society
Winning Student Paper
Experimental Evaluation of a MEMS Inertial Measurements Unit
for Doppler Navigation of Underwater Vehicles
Giancarlo Troni and Louis L. Whitcomb
Abstract— This paper reports the results of an in-water
laboratory experimental evaluation of the attitude estimation accuracy of a low-cost micro-electro-mechanical systems
(MEMS) attitude and heading reference system (AHRS), and
the effect of the accuracy of this sensor on Doppler-based
underwater navigation. We report a comparative analysis of
Doppler navigation obtained employing MEMS AHRS in comparison to Doppler navigation obtained with a high-accuracy
inertial navigation system (INS) including a true-North-seeking
gyrocompass and high precision accelerometers. The data indicate that Doppler navigation performance with MEMS AHRS
is sensitive to instrument calibration including Doppler/AHRS
alignment calibration, calibration of AHRS magnetometers for
hard-iron & soft-iron errors, and calibration of AHRS angular
rate sensors. When carefully calibrated, MEMS AHRS Doppler
navigation error is shown to be within an order-of-magnitude of
that obtained with high-end INS for the conditions and vehicle
trajectories studied. The goal of this evaluation is to quantify
Doppler navigation performance using MEMS AHRS. These
results may be useful in the development of lower-cost Doppler
navigation systems for small and low-cost underwater vehicles.
I. I NTRODUCTION
Bottom-lock Doppler sonar navigation is a common
method for high-precision near-bottom underwater vehicle
navigation. Doppler sonar navigation typically employs a
3-axis Doppler velocity log (DVL), a precision pressure
depth sensor, and a 3-axis attitude sensor [15]. High-end
North-seeking gyrocompasses are often employed to estimate
vehicle attitude [3], but advances in inertial measurement
technology have enabled the development of a new class of
compact, low power, low-cost attitude and heading reference
system (AHRS), [10].
This paper reports the results of a comparative experimental analysis of the performance of Doppler navigation using
low-cost micro-electro-mechanical systems (MEMS) AHRS
versus the performance of Doppler navigation using highend inertial navigation system (INS). Unlike most previous
studies which report numerical simulations, e.g. [2], we
report an in-water comparative experimental evaluation of
these navigation systems on the Johns Hopkins University
(JHU) remotely operated vehicle (ROV).
This paper is organized as follows: In Section II we give
a brief overview of Doppler navigation and the attitude
estimation. In Section III we describe our experimental setup
and evaluation methodology. In Section V we report the
This work was supported by the National Science Foundation under NSF
award IIS-0812138. Support for the first author was also provided by a
Fulbright/Conicyt Fellowship.
The authors are with the Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA. Email:
{gtroni,llw}@jhu.edu
978-1-4673-0831-1/12/$31.00 ©2012 IEEE
results of laboratory experiments to evaluate the performance
of the different attitude sensors. Section VI summarizes and
concludes.
Fig. 1.
JHU ROV inside the Johns Hopkins Hydrodynamic Test Facility
II. BACKGROUND
This section briefly reviews basic concepts of attitude
estimation and Doppler navigation of underwater vehicles.
A. Notation
For each vector, a leading superscript indicates the frame
of reference and a following subscript indicates the sensor
source, thus w pl is the LBL position in the world coordinates,
v
vd is the Doppler velocity sensor in the vehicle frame and
i
ai is the accelerometer linear acceleration in the inertial
sensor frame.
The set of 3×3 rotation matrices forms a group, known
as the special orthogonal group, SO(3), defined as
SO(3) = {R : R ∈ R3×3 , RT R = I, det(R) = 1}.
(1)
For each rotation matrix a leading superscript and subscript
indicates the frames of reference. For example, w
v R is the
rotation from the vehicle frame to the world frame.
B. Overview of Doppler Navigation
A Doppler sonar measures the vehicle’s three-axis velocity
in instrument coordinates with respect to the fixed sea
floor. The most common Doppler sonar configuration for
underwater vehicle navigation consists of four downwardlooking acoustic transducers each oriented at 30◦ from the
This is a DRAFT. As such it may not be cited in other works.
26IEEE Oceanic
Engineering Society Newsletter, January 2013
The citable Proceedings of the Conference will be published in
IEEE Xplore shortly after the conclusion of the conference.
instrument vertical axis. Each transducer measures acoustically the velocity of the instrument parallel to its beam with
respect to the fixed sea floor [8]. The instrument velocity in
the world frame sensor is given by
w
v d
vd (t) =w
v R(t) dR vd (t)
(2)
where vd R is the constant rotation matrix from Doppler
instrument coordinate frame to the vehicle coordinate frame,
and w
v R(t) is the time varying rotation matrix from the
vehicle coordinate frame to the inertial world coordinate
frame provided by an attitude sensor.
To obtain DVL position estimate, w pd (t), a common
solution use the dead-reckoning equation as follows [15]
w
pd (t) =w pd (t0 ) +
t
w
vd (τ ) dτ.
(3)
t0
MEMS AHRSs typically contain (i) a three-axis angular
rotation rate sensor, (ii) a three-axis linear accelerometer, and (iii) a three-axis magnetometer. Data from these
sensors are employed to estimate the the direction of the
earth’s gravitational field vector in instrument coordinates,
a = [ax , ay , az ]T . The estimated roll ϕ̂ and pitch θ̂ angles
[14] are given by
(4)
(5)
(6)
where ψ0 is the known local magnetic variation. Magnetic
distortions are commonly categorized as a hard iron and
soft iron effects [1]. Properly calibrated magnetometers can
provide heading accuracies on the order of 1◦ –3◦ with
respect to local magnetic North [6]. To improve the dynamic
performance of the AHRS attitude estimation, angular velocity gyrocompasses are used with a filter (such as the Kalman
filter or the complementary filter) to dynamically estimate
attitude.
High-end navigation-grade attitude sensors typically employ three-axis fiber-optic or ring-laser gyrocompasses to
estimate the Earth’s rotation and to estimate the direction
of true-North. Such sensors can yield dynamic heading
accuracies on the order of 0.1◦ and roll/pitch accuracy on
the order of 0.01◦ [4].
IEEE Oceanic Engineering Society Newsletter, January 2013
A. Experimental Setup
The facility contains a 7.5 m diameter × 4 m deep indoor
fresh water tank made of steel. The JHU ROV is actuated by
six 1.5 kW DC brushless electric thrusters and is capable of
being actively controlled in 6 degrees of freedom (DOF). A
suite of sensors commonly employed in deep submergence
underwater vehicles is present on the JHU ROV. Table I
details the JHU ROV attitude sensors. Table II specifies
the additional JHU ROV navigation sensors used in our
performance evaluation.
Instrument
High-end
INS
MEMS
AHRS
Model
IXSEA
PHINS III [4]
Variable
Specification Update Rate
Heading
0.1◦
Pitch/Roll
0.01◦
Angular Rate 0.01◦ /s (1 )
Acceleration
1 mg (1 )
Heading
Microstrain
Pitch/Roll
3DM-GX3-25 [6] Angular Rate
Acceleration
±2◦
±2◦
0.245◦ /s
0.65 mg
10 Hz
100 Hz
1:
Roll/pitch accuracies on the order of 0.1◦ (static) and 1◦ –5◦
(dynamic) are reported [6]. Three-axis flux-gate and magnetostrictive magnetometers are a common and inexpensive
sensors used to estimate heading. The measured magnetic
field i m ∈ R3 in the instrument frame can be transformed
from the instrument frame to the local-level frame by the
l i
l
3×3
is a rotation matrix
relation l m = R
i m, where iR ∈ R
using pitch and roll data estimates. Then the estimated truenorth heading ψ̂ [14] can be computed as
ψ̂ = atan2(−l my ,l mx ) − ψ0
The experimental evaluation reported herein used data
obtained with the JHU ROV, Figure 1, in the JHU Hydrodynamic Test Facility [5]. This Section gives an overview of
the experimental setup.
TABLE I
JHU ROV ATTITUDE S ENSORS P RECISION AND U PDATE R ATE
C. Overview of Attitude Sensors
ϕ̂ = atan2(−ay , −az )
θ̂ = atan2(ax , a2y + a2z ).
III. P ERFORMANCE E VALUATION
Phins raw output data is degraded to complain with exportation
regulations. Internally the sensors have a higher performance.
TABLE II
JHU ROV NAVIGATION S ENSORS P RECISION AND U PDATE R ATE
Instrument
Model
Variable
DVL
Teledyne RDI
1200 kHz [9]
Velocity
Pressure
sensor
Paroscientific [7]
Depth
LBL
Marquest
Sharps 300 kHz XY Position
LBL System
Specification Update Rate
±0.2%
±1 mm/s
8 Hz
0.01%
15 Hz
5 mm
5 Hz
The MEMS AHRS and high-end INS attitude data were resampled to the DVL sampling time to estimate the Doppler
navigation position. Then the estimated Doppler positions
were re-sampled to the LBL position to estimate the position
error. The Doppler velocity, the high-end INS attitude and
MEMS AHRS attitude are used in these experiments without
any extra post-processing or filtering. The LBL fix data
outliers were manually removed. In these experiments the
vehicle followed pre-programmed trajectories under closeloop control.
B. Evaluation Methodology
To analyze the DVL position estimation performance
based on the MEMS-based attitude sensor, we calculate the
following quantities:
27
a. Position error metric: Using the attitude estimation,
w
v R(t), from each sensor we recomputed the Doppler
track of the vehicle, w pd , using (3). We evaluated the
performance of each sensor by comparing the estimated
position with the “ground truth” LBL position, w pl , and
calculate the standard deviation,
w
σ̂ = [σˆx σˆx σˆx ] = σ(pl − pd )
(7)
and a position error metric (PE) is defined as
1
2
||σ ||2 = (σx2 + σy2 + σz2 ) .
(8)
b. Percentage of distance traveled: Using the standard
dead-reckoning navigation metric reporting the error of
the final position as a percentage of the distance traveled
(DT).
TABLE III
I NTERNAL S ENSORS N OISE P ERFORMANCE
Magnetometers
[mili-Gauss]
Accelerometers
[mili-g]
Angular Rate sensors
[deg/s]
MEMS AHRS - Microstrain 3DM-GX3-25
Meas.X
Meas.Y
Meas.Z
0.225
0.215
0.354
0.625
0.641
0.618
0.245
0.215
0.279
Specs
0.767
0.653
0.245
1.0001
0.0201
High-end INS - IXSEA Phins III
Specs
-
1 : Phins raw output data is degraded to complain with export regulations.
Internally the sensors have better performance.
IV. E XPERIMENTS AND R ESULTS
First a set of experiments was conducted to explore the
performance of each sensor required for attitude estimation.
Second, we compared the MEMS AHRS attitude estimation with that of the high-end INS. Third, an experiment
following a standard survey trajectory was conducted to
evaluate the influence of the attitude sensor performance on
the Doppler navigation position estimation.
A. Performance and Calibration of Internal Sensors
1) Static Noise: We analyzed the static noise characteristics of the internal sensors of the MEMS AHRS. Table IVA.1 shows the standard deviation of each sensor’s output data
while the vehicle was motionless for a period of 20 min. The
AHRS MEMS on-board sensors are internally filtered within
the AHRS. A user-selectable digital moving average filter is
applied to the sensor output. A digital window size of 66.7
Hz was used for the accelerometers data and angular rate
sensors, and a windows size of 58.8 Hz for the magnetometer. Observed noise is in agreement with the manufacturer’s
specifications, as shown in Table IV-A.1. The accelerometers
and angular rate sensors are exactly in the range specified,
and the magnetometers shows less noise than specified. For
the high-end INSs the specifications are artificially degraded
for export regulations, but internal inertial sensors in highend INSs have several order of magnitude better precision
than those in MEMS AHRSs.
2) Magnetometers: The magnetic disturbance due to the
presence of the 7.5 m diameter × 4 m deep steel water
tank degrades the heading estimation performance. Figure 2
shows the error in heading due to magnetic field distortions
inside the tank. We have identified a region of operation in
the upper half of the tank with more uniform magnetic field.
The experiments reported herein were conducted within this
region, thereby limiting the magnetic disturbances.
Calibration of the magnetometers sensors was performed
to remove the effect of hard and soft iron magnetic disturbances. Figure 3 shows the performances improvement in
heading estimation for the magnetic calibration for a specific
location in the center of the tank. The raw magnetic heading
measurement show more than 35◦ of heading error in this
Fig. 2. Tank heading disturbance effect: Heading error between the MEMS
AHRS and the high-end INS as reference. Colors show the heading error
in degrees for the sensor in different locations of the tank.
case. After a standard hard-iron calibration the heading error
was reduced to less than 5◦ . Then, after performing a softiron calibration, the heading error is reduced to less than 1◦
for this specific location in the center of the tank. For the
rest of this study this full hard-iron and soft-iron calibration
is used to improve the performance of the magnetometers.
3) Accelerometers: We compared the individual performance of the MEMS AHRS internal accelerometers to the
output from the internal accelerometers from the high-end
INS. The high-end INS specific unit used for this study
only reports the acceleration after the 1 G gravity vector
is removed. For this analysis the gravity vector is added in
post-processing. A calculated alignment calibration between
both attitude sensors is used to compare the acceleration.
To evaluate the performance of MEMS accelerometers an
experiment with sinusoidal vehicle motion in each degree
of freedom was implemented. Figure 4 shows an histogram
of the error between the high-end INS acceleration and
the MEMS AHRS internal accelerometers output data. The
error, similar to a normal distribution, shows a offset bias
less than 0.15 mg in each axis. The standard deviation is
approximately 0.6 mg in each axis.
28IEEE Oceanic Engineering Society Newsletter, January 2013
Relative Freq.[%]
HDGAHRS MAG−HDGINS[deg]
10
10
10
8
8
8
6
6
6
4
4
4
2
2
2
40
20
0
−20
0
−2
−40
0
50
100
150
HDG
200
[deg]
250
300
350
0
2
Ang.Vel.X[°/s]
0
−2
0
2
Ang.Vel.Y[°/s]
0
−2
0
2
Ang.Vel.Z[°/s]
INS
Uncalibrated
Hard Iron
Hard+Soft Iron
Fig. 5. Angular rate output comparison histogram. X-axis shows difference
between the MEMS AHRS and the high-end INS angular rate. Y-axis shows
the relative frequency of each acceleration error.
Relative Freq.[%]
Fig. 3.
Heading estimation performance after different magnetometer
calibration methods. X-axis shows the reference heading from the high-end
INS. Y-axis shows the relative heading error between the attitude sensors.
25
25
25
20
20
20
15
15
15
10
10
10
5
5
5
0
−0.2
0
0.2
Acc.X[m/s2]
0
−0.2
0
0.2
Acc.Y[m/s2]
0
−0.2
0
0.2
2
Acc.Z[m/s ]
Fig. 4.
Accelerometers output comparison histogram. X-axis shows
difference between the MEMS AHRS and the high-end INS acceleration.
Y-axis shows the relative frequency of each acceleration error.
4) Angular Rate Sensors: We compared the individual
performance of the MEMS AHRS angular rate sensors to
the output from the fiber-optic gyrocompasses in the highend INS. A calculated alignment calibration between both
attitude sensors is used to compare the angular rate sensors.
To evaluate the performance of the MEMS angular rate
sensors an experiment with sinusoidal vehicle motion in each
degree of freedom was implemented. Figure 5 shows an
histogram of the error between the high-end INS angular rate
and the MEMS AHRS internal angular rate sensors output
data. The standard deviation is less than 0.3◦ /s in each axis.
The normal distributed error shows a high component of bias
in several axis. The mean error for the angular rate error in
each axis is 0.682◦ /s in X, 0.015◦ /s in Y and -0.533◦ /s in
Z axis.
B. Attitude Estimation Performance
This section reports the attitude estimation performance of
the MEMS AHRS in comparison to the high-end navigationgrade INS. Heading, pitch and roll are evaluated under
standard underwater vehicle navigation conditions.
The accurate calibration of MEMS AHRS is critical to
get accurate results. Section IV-A.2 shows the performance
improvement in heading estimation due to an accurate calibration of the magnetometers.
IEEE Oceanic Engineering Society Newsletter, January 2013
The AHRS employed in this study, a Microstrain 3DMGX3-25 [6], estimates attitude (heading, pitch, and roll)
in real-time with a complementary filter employing data
from the unit’s 3-axis magnetometer, 3-axis accelerometer,
and its 3-axis angular rate sensors. Most MEMS angularrate sensors are affected by a bias offset that changes with
time and temperature and, in consequence, they are often
provided with a calibration procedure to correct the observed
angular-rate sensor bias. We observed with this model AHRS
that an uncorrected bias offset in the angular-rate sensors,
as shown in Figure 5, will cause a bias in the resulting
estimated attitude. Figure 6 shows the attitude estimated
in real-time by the AHRS’s complementary filter in the
presence of uncorrected angular-rate sensor bias and the
attitude estimated in post-processing using equations (4), (5)
and (6). These data show a mean difference between the
two attitude estimates of of +1.7 ◦ in heading, −0.3 ◦ in
pitch and −1.8 ◦ in roll. This internal AHRS bias error,
if uncorrected, is an extra source of error to the Doppler
navigation position estimation. In the following experiments,
the angular-rate sensor bias was corrected at the beginning
of each experiment by calculating a 30 second average of the
angular-rate sensor output while the vehicle was motionless.
Correcting the angular-rate sensor bias was observed to
reduce the AHRS reported attitude bias by an order-ofmagnitude.
A standard survey trajectory, shown in Figure 8, was
used to measure the performance of the MEMS AHRS filter
estimation output. The attitude error is measured comparing
the reported attitude from the MEMS AHRS and the highend INS. For comparison purposes, the attitude is estimated
from the magnetometers and accelerometers data without
using the filter with angular rate gyroscopes data using
equations (4), (5) and (6). Table IV shows the mean, peak-topeak (P2P), and standard deviation (STD) for attitude error
performance for the complete trajectory. Results shows that
the under these laboratory conditions with a high magnetic
disturbances the heading is highly affected with a mean of
1.7◦ , STD of 2.3◦ and P2P of 8.8◦ of error. Pitch and roll
estimation error is under 1◦ P2P. Figure 7 shows a section
of the attitude estimation error.
29
Hdg [deg]
95
the MEMS AHRS, and the high-end INS.
90
85
0
20
40
60
80
100
120
0
20
40
60
80
100
120
0
20
40
60
Time [sec]
80
100
120
Pitch [deg]
40
20
0
−20
−40
Roll [deg]
10
5
0
−5
−10
AHRS reported attitude
Attitude from raw data
Fig. 6.
AHRS estimated attitude vs the calculated attitude from the
magnetometers and accelerometers. X-axis shows the time in seconds. Yaxes show the attitude for the MEMS AHRS and the calculated attitude.
TABLE IV
ATTITUDE E STIMATION P ERFORMANCE
Fig. 8.
Error
Heading
[deg]
Error
Pitch
[deg]
Error
Roll
[deg]
MEMS AHRS
Filter Output
Mean
STD
P2P
1.668
2.261
8.880
0.013
0.125
1.018
-0.129
0.094
0.695
Calculated from
MEMS
AHRS
Mag. and Acc.
Mean
STD
P2P
1.603
2.397
15.403
-0.042
0.272
3.574
0.024
0.302
3.640
10
Hdg Error
[deg]
5
0
−5
−10
18
18.5
19
19.5
20
20.5
21
21.5
22
18.5
19
19.5
20
20.5
21
21.5
22
18.5
19
19.5
20
Time [min]
20.5
21
21.5
22
Pitch Error
[deg]
2
1
0
−1
−2
18
Roll Error
[deg]
2
1
0
−1
−2
18
MEMS AHRS
Calculated from Mag.and Acc.
Fig. 7. MEMS AHRS Attitude Estimation Error from high-end INS. X-axis
shows the time in minutes. Y-axes show the error attitude for the MEMS
AHRS and the calculated attitude.
C. Doppler Navigation Performance
This section reports the performance comparison of the
Doppler navigation using data from both attitude sensors,
LBL three-dimensional vehicle trajectory.
Although there are several sources of error that limits
Doppler navigation precision, previous studies have reported
that the accuracy of the calibration of the alignment between
the attitude sensor and the DVL can be a significant (and often dominant) source of navigation error, e.g. [11], [12]. Using the method reported in [13], the estimated DVL/MEMS
AHRS alignment calibration, vd R, was the rotation matrix
of +45.4◦ in heading, -1.1◦ in pitch, and +1.2◦ in roll.
Then based on the estimated alignment between both attitude
sensors the alignment matrix for the high-end INS and the
DVL sensor is calculated as +46.4◦ in heading, -0.4◦ in pitch,
and +0.2◦ in roll.
1) Doppler navigation based on each attitude sensor: We
analyze the performance of the Doppler navigation using data
from the MEMS AHRS and compared to the case using the
high-end INS, for the case of the vehicle following a standard
survey trajectory, shown in Figure 8. Each estimated position
is then compared with data from long-baseline (LBL) and
used as “ground truth” position. Figure 8 shows the JHU
ROV vehicle trajectory measured by the LBL. During the
35.8 min experiment the vehicle traveled 144 m.
Table V shows the position error for the complete trajectory and the percentage of the distance traveled for each
solution. Examination of the position error show that the best
performance is achieved by the high-end INS solution. The
estimated position error using the high-end INS is almost
three time smaller than the position error from the MEMS
AHRS. Figure 9 shows how the position error grows over
time for both solutions. Although the error depicted in Figure
9 correspond to a single realization of a random process,
the results show how a better attitude estimation reduce the
position drift over time.
2) Doppler navigation performance under different configuration scenarios: We analyze two test scenarios to
30IEEE Oceanic Engineering Society Newsletter, January 2013
TABLE V
S UMMARY OF THE D OPPLER NAVIGATION EVALUATION RESULTS
σx
σy
σz
High-end INS
0.062
0.048
0.021
MEMS AHRS
0.086
0.205
0.064
||
σ ||2
%DT XYZ
0.081
0.043
0.232
0.448
Pos X [m]
0.4
0.2
0
−0.2
−0.4
0
5
10
15
20
25
30
35
heading errors.
The second scenario, AHRS(S2), shows how the position
estimation performance decrease when using a less accurate
attitude estimation. As shown in Table IV, the magnetometers
calculated heading can be several degrees less accurate
than the heading estimated by the MEMS AHRS filter.
This less accurate estimated attitude makes the navigation
performance a 44% worse than the original case using the
attitude estimated by the MEMS AHRS filter. In all the cases
is important to notice that the performance is under 0.7% of
the distance traveled.
100
90
0
−0.5
0
5
10
15
20
25
30
35
Pos Z [m]
0.6
0.4
0.2
0
−0.2
Percentage Position Error (%)
Pos Y [m]
1
0.5
80
70
60
50
40
30
20
10
0
5
10
High−end INS
15
20
Time [min]
25
30
35
0
0
0.05
0.1
MEMS AHRS
INS
Fig. 9. Doppler navigation estimated position error. X-axis shows the time
in minutes. Y-axes show the error for each calculated position.
evaluate the Doppler navigation performance under different
conditions:
a. AHRS(S1): This scenario simulates the case of having
accurate heading without magnetic disturbances. In this
case, to estimate the Doppler navigation position, we
use the high-end INS reported heading after adding
a hard-iron disturbance equivalent to ±1◦ of heading
error. Pitch and roll are from the MEMS AHRS.
b. AHRS(S2): This second scenario simulates the case of
not having an attitude filter implemented or angular rate
data available. In this case we used the low-pass filtered
data from the magnetometers and accelerometers and
equations (4), (5) and (6) to calculate the attitude used
to estimate the Doppler navigation position.
Figure 10 and 11 show the results for the Doppler navigation performance under the two defined scenarios, and
also the basic cases reported in previous section. Figure 10
shows the position error distribution. Figure 11 summarize
the position error and the percentage of the distance traveled
for all the scenarios defined.
Examination of the position error shows that the scenario AHRS(S1) improve the navigation performance by
35% compared to the case using of the MEMS AHRS.
The position error performance is only two times worse
than the best case using of the high-end INS. This case
should represent the best expected performance under this
circumstances for Doppler position navigation based on the
MEMS AHRS. Also scenario AHRS(S1) shows that for this
analyzed trajectory Doppler navigation is very sensitive to
IEEE Oceanic Engineering Society Newsletter, January 2013
0.15
0.2
0.25
Position Error [m]
AHRS
AHRS(S1)
0.3
0.35
0.4
AHRS(S2)
Fig. 10. Position Error Distribution. The y-axis shows the percentage of
the XYZ position error that was under certain threshold (x-axis).
Fig. 11. Summary of the Doppler navigation evaluation results under
different scenarios. Left figure shows the potion error performance (x-axis).
Right figure shows the final position error as a percentage of the distance
traveled (x-axis).
V. C ONCLUSIONS
We conclude the following from our comparative in-water
experimental evaluation of underwater Doppler navigation
with a MEMS AHRS in comparison with a high-end INS.
Accurate MEMS AHRS instrument calibration is required to obtain fair Doppler navigation performance. Results
showed that calibration of AHRS magnetometers for hardiron & soft-iron errors highly improves the heading estimation, and thereby improves Doppler navigation performance.
Calibration of AHRS angular rate sensors offset degraded
the reported attitude estimation and consequently Doppler
31
navigation performance. Also Doppler/AHRS alignment calibration is a significant source of Doppler navigation error.
Results for the conditions and vehicle trajectories studied
allows us to quantify the position estimation performance
under different scenarios commonly found in underwater
Doppler navigation. Doppler navigation error, based on a
carefully calibrated MEMS AHRS, is shown to be within an
order-of-magnitude of that obtained with high-end INS. Also
for the trajectories studied, Doppler navigation shows to be
very sensitive to errors in MEMS AHRS heading estimation.
Finally, MEMS AHRS attitude estimation filter shows to
improve the Doppler navigation performance within a third
of that obtained with the calculated attitude only based on
the AHRS on-board magnetometers and accelerometers.
These results may be useful for better understand the
performance of Doppler navigation systems under different
configurations and ultimately lower the cost of underwater
vehicles.
R EFERENCES
[1] M. Caruso. Applications of magnetic sensors for low cost compass
systems. In Position Location and Navigation Symposium, IEEE 2000,
pages 177 –184, 2000.
[2] Y. Geng, R. Martins, and J. Sousa.
Accuracy analysis of
DVL/IMU/magnetometer integrated navigation system using different
IMUs in AUV. In Control and Automation (ICCA), 2010 8th IEEE
International Conference on, pages 516 –521, June 2010.
[3] C. R. German, A. Bowen, M. L. Coleman, D. L. Honig, J. A. Huber,
M. V. Jakuba, J. C. Kinsey, M. D. Kurz, S. Leroy, J. M. McDermott,
B. M. de Lépinay, K. Nakamura, J. S. Seewald, J. L. Smith, S. P. Sylva,
C. L. Van Dover, L. L. Whitcomb, and D. R. Yoerger. Diverse styles
of submarine venting on the ultraslow spreading Mid-Cayman Rise.
Proceedings of the National Academy of Sciences, 107(32):14020–
14025, 2010.
[4] IXSEA. PHINS III User Guide. IXSEA, MU-PHINSIII-002 A edition,
July 2008.
[5] J. C. Kinsey, D. A. Smallwood, and L. L. Whitcomb. A new
hydrodynamics test facility for UUV dynamics and control research.
In OCEANS 2003, volume 1, pages 356–361 Vol.1, 2003.
[6] Microstrain Inc. 3DM-GX3-25 miniature attitude heading reference
system datasheet. Technical Report P/N 8400-0033 rev. 002, Microstrain Inc., Williston, VT, 2012.
[7] Paroscientific Inc. Submersible depth sensors series 8000. Technical
report, Paroscientific Inc., Redmond, WA, 2005.
[8] RD Instruments Inc. Acoustic Doppler current profilers principals of
operation: A practical primer (2nd edition). Technical Report P/N
951-6069-00, RD Instruments Inc, San Diego, CA, 1996.
[9] Teledyne RD Instruments Inc. Workhorse navigator Doppler velocity
log (DVL). Technical report, RD Instruments Inc, Poway, CA, 2011.
[10] D. Titterton and J. Weston. Strapdown Inertial Navigation Technology,
Second Edition (Progress in Astronautics and Aeronautics). AIAA,
2nd edition, 2004.
[11] G. Troni, J. Kinsey, D. Yoerger, and L. Whitcomb. Field performance
evaluation of new methods for in-situ calibration of attitude and
Doppler sensors for underwater vehicle navigation. In Robotics and
Automation (ICRA), 2012 IEEE International Conference on, pages
5334 –5339, may 2012.
[12] G. Troni and L. L. Whitcomb. Experimental evaluation of new
methods for in-situ calibration of attitude and Doppler sensors for
underwater vehicle navigation. In Intelligent Robots and Systems
(IROS), 2011 IEEE/RSJ International Conference on, pages 3722 –
3727, sept. 2011.
[13] G. Troni and L. L. Whitcomb. Preliminary experimental evaluation
of in-situ calibration methods of MEMS-based attitude sensors and
Doppler sonars for underwater vehicle navigation. Autonomous Underwater Vehicles (AUV), 2012 IEEE/OES, september 2012. To Appear.
[14] M. Wang, Y. Yang, R. Hatch, and Y. Zhang. Adaptive filter for a
miniature MEMS based attitude and heading reference system. In
Position Location and Navigation Symposium, 2004. PLANS 2004,
pages 193 – 200, april 2004.
[15] L. L. Whitcomb, D. Yoerger, and H. Singh. Advances in Dopplerbased navigation of underwater robotic vehicles. In IEEE Int. Conf.
on Robotics and Automation., volume 1, pages 399–406 vol.1, 1999.
32IEEE Oceanic Engineering Society Newsletter, January 2013
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement