november 2003 kc-135 spheres flight test results
DESCRIPTION
November 2003 KC-135 SPHERES flight test results. Mark O. Hilstad, Simon Nolet, Dustin Berkovitz, Alvar Saenz-Otero, Dr. Edmund Kong, and Prof. David W. Miller MIT Space Systems Laboratory 2003-11-24. Overview. Attitude control and beacon tracking Tracking of a hand-held beacon by a sphere. - PowerPoint PPT PresentationTRANSCRIPT
November 2003 KC-135 SPHERES flight test results
Mark O. Hilstad, Simon Nolet, Dustin Berkovitz, Alvar Saenz-Otero, Dr. Edmund Kong, and Prof. David W. Miller
MIT Space Systems Laboratory2003-11-24
Nov. 2003 KC-135 results 2
Overview
• Attitude control and beacon tracking– Tracking of a hand-held beacon by a sphere.– Tracking of a sphere by another sphere.
• Simple search pattern– Open-loop three-axis rotation
• Coordinated search pattern– Elements of the “lost in space” maneuver
• Docking– Initial approach stage only
• Identification of inertia and center of mass• Lessons learned
A series of four KC-135 flights during the week of 3 Nov 2003 was sponsored by the Jet Propulsion Laboratory’s Terrestrial Planet Finder program. These flights were used to test algorithms designed by several members of the SPHERES team, and by NASA Ames. The results of these tests are presented herein.
This presentation referencesthe following video files:
• KC135_Nov03_flight2_para23.mpg• KC135_Nov03_flight4_para02.mpg• KC135_Nov03_flight4_para03.mpg• KC135_Nov03_flight4_para10.mpg• KC135_Nov03_flight4_para13.mpg• KC135_Nov03_flight4_para18.mpg• KC135_Nov03_flight4_para19.mpg• KC135_Nov03_flight4_para29.mpg
Nov. 2003 KC-135 results 3
Beacon tracking experiments
• Objective– Demonstrate tracking of a beacon
and optimal rotation along the shortest path
– Validate the 3-D control law by demonstrating off-axis 3-D attitude control
• Experiment description– Tracking of a beacon randomly
located in the test volume
– Tracking of a free floating sphere’s on-board beacon
– Two spheres simultaneously tracking each others’ on-board beacons
Flight 2, parabola 23
Flight 4, parabola 18
Nov. 2003 KC-135 results 4
Beacon tracking results
• Flight 4, parabola 18– Pointing error is reduced in each test (q1, q2 and q3 tend toward zero, while
q4 tends toward one)
– Body rates show angular acceleration and deceleration, as expected
Test 1 Test 2Test 1 Test 2 Test 1 Test 2
Nov. 2003 KC-135 results 5
Simple searchF
light 4, parabola 2F
light 4, parabola 3
• Initial conditions– A single sphere stationary with respect
to the KC frame
• Open-loop spin– Try to point the onboard beacon in as
many directions as possible• Attempt to map 4 steradians
• Alternating thrusters – Two thrusters on at any given time
• Very limited time, so maximize actuation
– Used to change the plane of rotation– Thruster state changes once per
second.
• A propellant-efficient algorithm will be used for SPHERES-TPF.
Nov. 2003 KC-135 results 6
Simple search results
• Day 4, parabola 2– Alternating z, y, x torque commanded– Some coverage due to thrusters, some due to disturbances such as bumping walls– Sparse coverage in the -x direction, but within the half-cone angle of the beacon– Quaternion integration issues add uncertainty to validity of results
• Quaternion behavior appears erratic, but rapid change is expected• Rate gyroscope saturation leads to inaccurate quaternion integration• Quaternion normalization errors also led to inaccurate integration
– Video shows multi-axis rotation, as intended.
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1P
os
itio
n [
m]
Sphere 1 state telemetry.
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Ve
loc
ity
[m
/s]
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Qu
ate
rnio
n
0 2 4 6 8 10 12 14 16-2
-1
0
1
2
Time [s]
Bo
dy
ra
te [
rad
/s]
xyz
xyz
q1
q2
q3
q
xyz
Coverage maps
-10
1
-1
01-1
0
1
XY
Z
-1 0 1-1
-0.5
0
0.5
1
X
Y
-1 0 1-1
-0.5
0
0.5
1
X
Z
-101-1
-0.5
0
0.5
1
Y
Z
Nov. 2003 KC-135 results 7
Coordinated search
• Experiment description– Two spheres begin with the
beacon faces offset from their common line of sight by 135°.
– Both spheres initiate open-loop z-axis spins.
– Acquisition occurs when one sphere hears the other’s beacon.
– When a sphere hears a beacon, it sends a stop message to the other sphere.
– When a sphere receives a stop message, it initiates rate damping.
– Both spheres use the same algorithm.
Flight 4, parabola 10
Flight 4, parabola 13
Nov. 2003 KC-135 results 8
Coordinated search results
• Flight 4, parabola 10– Local 0° are offset from common line by -135°– Perform +z spin at ~30°/s until acquisition/communication, followed by rate damping
• Beacon half angle is ~30-45°(equivalent to 1-1.5 seconds of spin time)– Telemetry from both spheres shows expected behavior
• Sphere 1 shows rate damping at z-quat≈0.6 → z-angle≈106°• Sphere 2 shows rate damping at z-quat≈0.45 → z-angle≈127°
– Primary maneuver (spin search, then decelerate upon acquisition) is circled.
Sphere 1
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Po
sit
ion
[m
]
Sphere 1 state telemetry.
xyz
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Ve
loc
ity
[m
/s]
xyz
0 2 4 6 8 10 12 14 16-0.5
0
0.5
1
Qu
ate
rnio
n
q1
q2
q3
q
0 2 4 6 8 10 12 14 16-2
-1
0
1
2
Time [s]
Bo
dy
ra
te [
rad
/s]
xyz
Rate damping begins
q3 at start of rate damping
Sphere 2
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Po
sit
ion
[m
]
Sphere 2 state telemetry.
xyz
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Ve
loc
ity
[m
/s]
xyz
0 2 4 6 8 10 12 14 16-1
-0.5
0
0.5
1
Qu
ate
rnio
n
q1
q2
q3
q
0 2 4 6 8 10 12 14 16-2
-1
0
1
2
Time [s]
Bo
dy
ra
te [
rad
/s]
xyz
Rate damping begins
q3 at start of rate damping
Nov. 2003 KC-135 results 9
Glide slope docking experimentsF
light 4, parabola 19F
light 4, parabola 29
• Objective– Demonstrate the first phase of the
docking approach computed by a glide slope algorithm
• Experiment description– The two spheres maintain their
orientation toward each other
– The chasing sphere initiates a translation along its x-body axis to move toward the target
– The algorithm is set such that docking should occur in about 8 seconds!
Nov. 2003 KC-135 results 10
Glide slope docking results
• Flight 4, parabola 19– The first 3.5 seconds of the docking maneuver were successfully achieved
– The spheres maintain relative pointing, as shown by the constant quaternions
– Similar data were acquired by both spheres using their own sensors
Docking Approach
Range to target
Approach velocity
Chasing Sphere
Docking Approach
Range to target
Approach velocity
Chasing Sphere Range
to target
Approach velocity
Target Sphere
Docking Approach
Nov. 2003 KC-135 results 11
Inertial property identification
• Online gyro-based mass property identification– Use gyro data and calibrated thruster information to identify the error from nominal
values of center of mass offset and moment of inertia– Algorithms by SPHERES team and NASA Ames
Dry inertia
(kg m2)
Ixx0.022901
Iyy0.020546
Izz0.018165
Ixy0.000196
Ixz-0.000066
Iyz-0.000212
CM offset
(mm)
Full tank
CMx-0.054495248
CMy-0.810686405
CMz0.244119942
Empty tank
CMx-0.015665375
CMy-0.821387572
CMz3.081773236
Pre-flight estimated inertial properties• KC-135 flight goal: perform one
long test over many parabolas– Allows more time for estimates
to converge– Provides the most amount of
raw data for download
• Problem: disturbances corrupt identification
– Update stops by itself when gyro saturates, no thrusters are firing, or SNR is too low
– Pause/resume command used when sphere is handled
Nov. 2003 KC-135 results 12
Inertia ID sample data• Estimated angular acceleration (using rate sensors)
– Data are taken from parabolas 32-40, flight #4– Green lines indicate when test was paused– Delays in pausing and unpredictable motion caused issues
• Parabola 35: attached proof mass (a spare battery pack) on the -x face of the sphere
• Parabola 37: replaced the full tank with one containing only 28g of gaseous CO2 (no liquid) to remove effects of propellant slosh
• Parabola 38: removed the proof mass 0 10 20 30 40 50 60 70 80
-1
-0.5
0
0.5
1
Filtered angular accelerations, clean data only
time [sec]
alp
ha [
rad
/sec2 ]
ax
ay
az
Par
abol
a 35
Par
abol
a 37
Par
abol
a 38
ay and az dropWhen proof massIs attached
Nov. 2003 KC-135 results 13
Inertia ID results
• Deviations from the previous best-estimate inertia – These are “dry” values; propellant is subtracted out
• The inertia in the y (green) and z (red) axes jumps when the proof mass is added– The change is about 80% of what was expected; this is under investigation
• Propellant slosh is not an issue
0 10 20 30 40 50 60 70 80
-1
-0.5
0
0.5
1
1.5
2
2.5
x 10-3 ID'ed Inertia
time [seconds]
Inert
ia [
kg
-m2 ]
IxxIyyIzzIxyIxzIyzbetween parabolasproof mass added28g tank replacedproof mass removed
~ 3.2e-3 change(expected 3.73e-3)
Nov. 2003 KC-135 results 14
Center of mass ID results
• Deviations from the nominal CM offset. – These are “dry” values; propellant is subtracted out
• Since it was impossible to filter out all handling of the sphere using pause and resume commands in real-time, the online estimate was corrupted
• Saved high-frequency IMU data from the flight were used successfully in an identical off-line algorithm (accurate to floating-point precision) to produce promising results for both inertial and CM estimation.
0 10 20 30 40 50 60 70 80
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
ID'ed deltaC
time [seconds]
CM
off
set
[mm
]
deltaCx
deltaCy
deltaCz
Pa
rab
ola
35
Pa
rab
ola
37
Pa
rab
ola
38~ 5 mm change
(expected 6.3)
Nov. 2003 KC-135 results 15
Lessons learned
• Care must be taken when integrating separate tests into one program– Tests being integrated in one program must have compatible initialization settings– Special attention is required when the tests are written by different individuals
• Unexpected infrared noise causes confusing behavior– We have identified a series of fixes for use in any future KC flights
• IR-opaque, visible-transparent sheeting on lights (will also improve video quality)• Form-fitted, pre-cut curtains to block IR from other experiments• Point laptop screen away from test area• We expect these fixes to fully alleviate all infrared problems
– Recent test results on the ISS suggest that infrared noise will not be a problem; however, experiences like this continue to help us better understand noise sources and to identify new ways to address the issue if it arises.
• SPHERES core code changes– Static variables must be explicitly initialized at the start of each test, not just at the
start of the program– Order of process initialization problems manifested on flights as corrupt IMU data
• Identified problem and fixed by changing initialization procedure– Will add data to the telemetry stream to notify us of suspected IR noise.
Nov. 2003 KC-135 results 16
Conclusion
• Objectives accomplished– Showed that the attitude control algorithm works as
expected in 3D– Validated critical maneuvers and demonstrated key
components of the “lost in space” sequence• Open-loop search• Beacon acquisition• Intercommunication• Stop and hold response• Beacon tracking
• Saved precious experiment time on ISS– Gained confidence in current algorithms, and identified
areas for improvement prior to flight.– Improved the inertia and center of mass estimates of the
spheres.