University of Maryland Concepts and Technologies for Robotic Servicing of Hubble Space Telescope

1
University of Maryland Concepts and Technologies
for Robotic Servicing of Hubble Space Telescope

• David Akin, Brian Roberts,
• Walt Smith, and Brook Sullivan
• Space Systems Laboratory
• University of Maryland,College Park

2
Presentation Overview

• Space Systems Laboratory background
• Relevant SSL technologies
• Ranger system and experiences
• Recent HST studies
• Mission concepts
• Conclusions

3
ARAMIS Telerobotics Study

• Survey of five NASA Great Observatories to
  assess impacts and benefits of telerobotic
  servicing - major results
• Ground-controlled telerobotics is a pivotal
  technology for future space operations
• Robotic system should be designed to perform
  EVA-equivalent tasks using EVA interfaces
• Maximum market penetration for robot
• Maximum operational reliability
• Designing to EVA standards well understood
• Fully capable robotic system needs to be able to
  do rendezvous and proximity operations, grapple,
  dexterous manipulation

4
Fundamental Concept of Robotic Servicing
5
Beam Assembly Teleoperator
6
SSL Relevant Experience Timeline (1)
80
81
82
83
84
85
86
87
88
89
SSL studies applications of automation, robotics,
and machine intelligence for servicing Hubble and
other Great Observatories for NASA MSFC
BAT used for extensive servicing tests on HST
training mockup
Initial operational tests of Beam Assembly
Teleoperator
SSL develops ParaShield flight test vehicle for
suborbital mission
Experimental Assembly of Structures in EVA flies
on STS 61-B
7
Ranger Telerobotic Flight Experiment
8
SSL Relevant Experience Timeline (2)
90
91
92
93
94
95
96
97
98
99
Ranger performs end-to-end HST servicing
simulations
UMd NBRF opens
RTSX PDR
RTSX CDR
SSL designs Ranger based on experience with HST
servicing
Phase 0 PSRP
Phase 1 PSRP
Phase 2 PSRP
Ranger NBV operational
Environmental testing at JSC
NASA selects Ranger TFX as low-cost robotic
flight experiment
SSL directed to redesign Ranger for shuttle
mission Ranger TSX
9
Ranger Neutral Buoyancy Vehicle I
10
Ranger Telerobotic Shuttle Experiment
11
SSL Relevant Experience Timeline (3)
2002
2003
2004
2000
2001
Ranger TSX program cancelled
Modular miniature servicer development for DARPA
Development of ECU operations timeline
Dual-arm system in active test
All-up mockup for public outreach
PXL in NB testing
12
Robotic HST Servicing - Batteries
RANGER (2003)
BAT (1987)
13
Robotic HST Servicing - Instruments
ECU
WFPC
FGS
14
Ranger Flight Dexterous Arms
15
Dexterous Arm Design Objectives

• Approximately EVA-glove-sized end effectors with
  30 lbf capability and 30 lb-ft torque capability
  in any direction, and 45 per second joint
  velocities
• Allow safe surface contact despite communications
  time delays requires active compliance-control
  loop closed onboard
• Sufficient arm articulation for a wide range of
  tasks in a cluttered workspace
• Allow exchange of specialized end effectors
• Two mechanical tool drives to each end effector
• For cooperative tasks, two arms with intersecting
  workspaces mounted on narrow base
• Where feasible, additional sensors as alignment
  aids

16
Dexterous Manipulator System Design
Objectives
Design Consequences
Performance goals --- of sensors, actuators ---
of wires Envelope requirements Reliability ---
No external wiring
Serial communications, constant-length wiring
through joint axes
Local processors for inner control loops
Short wiring runs for analog signals
Internal temperature monitoring Force/Torque
sensing requirement
Component temperature limits --- thermal analysis
results Reliability, low maintenance Performance,
low friction
Brushless DC motors for arm actuation
Internal heat sinks Aluminum and copper
circumferential conductive paths
Thermal environment Component temperature limits
--- thermal analysis results
17
Dexterous Manip System Design (cont.)
Objectives
Design Consequences
Current driver circuits rather than Voltage
drivers
Electrical Power limitations Thermal variation
Unregulated Bus for high efficiency
Thermal environment 45-day nominal active design
life without access for recalibration
Joint-level control loops based on digital
sensors to avoid analog drift and noise
Co-located position sensors for high-gain inner
joint loops
Control bandwidth goals Friction rejection
Large harmonic drives
Compliance Loop performance requirements ---
stiff drive, no backlash
Motors inside harmonic drives
Envelope requirements
3 DOF intersecting axis shoulder 4 DOF
intersecting axis wrist 1 elbow pitch w/offset
Compliance Loop performance requirements ---
high outer loop rate --- simple
kinematics Contact stability requirements Task
requirements --- 7 DOF (w/elbow pose) Envelope
requirements Workspace requirements
18
4-Axis Skew Wrist Design
19
Why a Skew Axis?
Lower interference of the tool with the forearm
extends pitch travel Single-sided support of
the inner wrist allows for greater yaw range
Frontal area of the wrist reduced by skew layback
of pitch actuator
Skew 4-axis
Orthogonal 4-axis
20
Wrist Workspace Evolution
21
Toolspace Comparison
f (deg)
22
Dexterous Arm Cross-Section
23
Inner Wrist (Exploded View)
24
Bearing/Housing Design

• Problem Steel (a 7 µin/inF) bearing
  installed in Aluminum (a 13.1 µin/inF)
  housing. Over a wide temperature range the
  mismatch can cause the bearing to spin in housing
  or seize.
• Solution Install appropriately-sized Invar (a
  0.7 µin/inF) sleeve between bearing and housing.
  Sleeve size is chosen taking into account
  elastic deformation of bearing race, sleeve, and
  housing. Preload is provided by interference
  (shrink) fit of sleeve in housing.

25
Bearing Design Example
Sleeve Installed Between Bearing and Housing
Bearing Installed Directly in Housing
Effect of Invar sleeve is to make bearing preload
invariate with temperature
26
Actuator Performance Summary
27
Dexterous Arm Parameters

• Modular arm with co-located electronics
• Embedded 386EX rad-tolerant processors
• Only power and 1553 data passed along arm
• 53 inch reach mounting plate-tool interface plate
• 8 DOF with two additional tool drives (10
  actuators)
• Interchangeable end effectors with secure tool
  exchange
• 30 pounds tip force, full extension
• 150 pounds (could be significantly reduced)
• 250 W (average 1G ops)

28
Design Loads

• Nominal Operations Structural Load Path
• Fx, Fy, Fz 300 lbf
• Mx, My, Mz 200 lb-ft
• Actuator Input-side Loads (pure moment)
• Shoulder Actuators 650 oz-in
• Elbow, Wrist Roll Actuators 400 oz-in
• Wrist Pitch, Yaw, Hand Roll, Slow Tool Actuators
  230 oz-in
• Fast Tool Actuator 240 oz-in
• Actuator Output-side Loads (pure moment)
• Shoulder Actuators 200 lb-ft
• Elbow Actuators 120 lb-ft
• Wrist Roll Actuator 70 lb-ft
• Wrist Pitch, Yaw, Hand Roll, Slow Tool Actuators
  50 lb-ft

29
Design Loads (continued)

• Collision Loads External Housings
• Maximal collision 10 Joules (equivalent to 20
  kg _at_ 1 meter/sec)
• Method assume kinetic energy converted to
  strain energy find stresses for concentrated
  loading of component that produces the same
  strain energy
• Launch and Landing Loads
• Assume Stowed Configuration
• Accelerations
• X 7.9 G
• Y 4.9 G
• Z 8.3 G, - 6.3 G
• Subject to refinement with coupled-loads analysis
• Consider additional loads in some cases
• Thermal ambient operational limits -60 to
  100 C
• Assembly

30
Generalized Inverse Kinematics
HAND CONTROLLER
WRIST JOINT ANGLES
f,?,?
0
0
R
R
T
T
des
WRIST FORWARD KINEMATICS
DESIRED TOOL ORIENATION
ROTATIONAL CHANGE
?r

??
J
WRIST PSEUDOINVERSE JACOBIAN
p
W
WRIST JACOBIAN
J
W

0
/
??
J

0
/
W
WRIST NULLSPACE JACOBIAN
r
0
/
MANIPULABILITY, JOINT LIMIT INDEX
NULLSPACE VELOCITY
??
W
31
Singularity Avoidance (Experiment)
013
Forearm rolls to avoid singularity
32
Interchangeable End Effector Mech.

• 3 Mechanical Interfaces
• Hand Roll Drive
• Fast Tool Drive
• Slow Tool Drive
• No power or data interface

Each IEEM is approximately 2.75 Ø by 2. Weight
is 2 lbs.
33
Interchangeable End-Effector Mechanism (IEEM)

• Objectives
• Permit teleoperated and possible autonomous tool
  changeout
• Allow a rigid connection of Hand Roll DOF to tool
  structure
• Allow a connection of two, Tool Drive actuators
• Avoid introducing extra actuators or sensors
• Minimize risk of tool detaching from DX
  Manipulator
• Minimize risk of tool detaching from Tool Post
• Package mechanism compactly
• Minimize Tool/Wrist-center distance
• Maintain clear Wrist Camera field of view
• Maintain center shaft clear for boresight laser
• Design Consequences 5-State Mechanism
• IEEM locked to Tool Post, back-driving blocked
• DX Manipulator Soft-Docked to IEEM, IEEM locked
  to Tool Post
• IEEM locked to both DX Manipulator and Tool Post
• IEEM locked to DX Manipulator, Soft-Docked with
  Tool Post
• IEEM locked to DX Manipulator, free from Tool
  Post, back-driving blocked

34
IEEM Attached to Tool Post
35
IEEM Exploded View
36
State 1 IEEM Locked to Tool Post

• Slotted Ring retains mushroom end of Tool Post
• Two Ball-locks prevent ring rotation

37
State 2 Manipulator Engages IEEM

• Three teeth on Manipulator engage Slotted Ring
• Two Ball-locks are released, permitting Ring
  rotation
• Tool drives self-align and engage

38
States 3, 4 Manipulator Rotates 60

• State 3 Three flanges on Manipulator engage
  flanges on IEEM at beginning of rotation, locking
  Manipulator axially to IEEM
• State 4 At end of rotation, mushroom end of
  Tool Post is free to disengage from IEEM

39
State 5 Disengage Tool Post

• Outward radial motion of Tool Post rotates
  Blocking Latch, preventing reverse rotation of
  Slotted Ring
• When Tool Post is free, spring plunger in detent
  prevents back-rotation of Blocking Latch

40
IEEM Bearing Design

• Problem Locking Collar has poor thermal contact
  with IEEM inner structure, since it is supported
  on ball bearings. If a large temperature
  difference develops, the bearings could seize.
• Solution Design contact angle of bearings such
  that contact cones intersect on centerline of
  IEEM. Then any differential expansion of Locking
  Collar w.r.t. center structure occurs parallel to
  bearing contact, not affecting clearance or
  preload.

41
Schematic of IEEM Bearings
42
Wrist Camera View
IEEM
Tool Side
Arm Side
43
Tool Drives

• Tool Drive Motor Controllers are primary method
  for commanding / sensing EE gripping force or
  output torque
• Tool Drive Motor Specifications
• Hand Roll Drive (High Torque, Low Speed)
• Slow Tool Drive (High Torque, Low Speed)
• 52 ft-lbs, 139 /s no load
• Fast Tool Drive (Low Torque, High Speed)
• 1 ft-lb, 15,675 /s no load
• Must add gearing to use

44
End Effector Design Drivers

• Structural / Mechanical Requirements
• Grasp must be non-back-driveable
• Factor of safety of 1.4 (Test) or 2.0 ( Analysis)
• Withstand 300 lbs Force Along All Axes
• Withstand 200 ft-lbs Moment About All Axes
• Survive 6.75J Impact
• Withstand Launch Landing Loads
• Meet NASA EVA contact spec
• Meet RTSX Thermal Range
• Operational
• Visual Turn Indicators
• Tool Color Should Contrast With Task Equipment
  (Not Black)
• Visual Access for Grasp Verification
• Neutral Buoyancy Compatible

45
RTSX End Effectors
Microconical End Effector
Bare Bolt Drive
Right Angle Drive
Tether Loop Gripper
EVA Handrail Gripper
SPAR Gripper
46
Bare Bolt Drive (BBD)
Interfaces 7/16 Hex Head Bolts ECU Keyway
Slot Bolts APFR Latch Bolts Characteristics
5 lbs. 12 x 3ø Status NBV version in use
NBV Prototype
IEEM
47
Microconical End Effector (MEE)
Interfaces RPCM Microconical Interface
Characteristics 12 lbs. est. 8 x 3ø
Status Drawings and Stress Analysis for EVA
version received from OSS IDEAS models
completed Not currently fabricated
Microconical Interface
IEEM
48
Right Angle Drive (RAD)
Interfaces 7/16 Hex Head Bolts ECU
Connector Drive Characteristics 6 lbs. 8 x
6 x 3 Status NBV version in use
NBV Prototype
IEEM
49
APFR Paddle Gripper (APG)

• Interfaces
• APFR Locking Collar
• APFR Paddles
• APFR Detent Levers
• Characteristics
• 9 Lbs.
• 14 X 11 X 3
• PJM Based
• Typical of parallel jaw mechanism flexibility for
  specialized interfaces

Levers
Collar
Paddles
IEEM
50
Task Interfaces - ECU
Electronic Controller Unit
Tether Loop Gripper
BareBolt Drive
Right Angle Drive
51
Task Interfaces - APFR
Yoke
Interfaces Include 1. Quarter Turn Rings (2) 2.
Latch Bolts (2) 3. Lock Knob 4. Yoke 5. Detent
Levers 6. Locking Collar 7. Paddles
AQP BBD AQP TLG APG APG APG
Plus use BBD to rotate APFR probe
52
Ranger PXL System Requirements

• Device to supply full 6 DOF mobility within a
  limited work volume
• Forcemax 25 lbf Torquemax 200 ft-lbf
• Stiffness 250 Klbf/rad Angular Speedmax
  12º/sec
• End point accuracy /-.12, precision .01
• Provide means to react loads generated by Ranger
  performing task duties while minimizing overall
  deflection
• Design to maximize natural frequency of device
  with a target minimum of 10 Hz
• Capable of reacting 580 ft-lbf torque due to
  Primary Reaction Control System firings (PRCS)
• Brakes engage when no power is supplied
• Brakes must have manual (EVA) over-ride
• EVA interfaces to be at least 24 away from SLP
  end

53
Design Requirements

• Control electronics same as in the DX VM
• Co-located at actuator operated by serial command
  1553
• Main power 48VDC 10A max, control power 28VDC
• No changing form factors on cards from DX design
• Pass through hole to be integrated in spine of
  mechanism approximately 1 diameter
• Drive system to use DC brush-less motors
• Each actuator to have triple redundant position
  feedback. One feedback system gives absolute
  position information that maintains position
  after power loss
• Actuator drive system back drive-able under zero
  power conditions
• Actuators connected using aerospace standard
  V-couplings to facilitate assembly

54
DesignPXL Assembly
Electronics Housing
EVA Interface
Ø 9.5
75
Roll Joints
Pitch Joints
19
55
Design Roll Joint
Main Drive Gear
Motor
Outer Housing
EVA Interface
Inner Housing
Brake Release
Brake
Wave Generator
Driveshaft
Incremental Encoder
Pass Through Tube (Ø .80)
Flexspline
Absolute Encoder
56
PXL in Stowed Configuration
Side View
57
Components Harmonic Drive

• Double the torsional stiffness
• Double the peak torque ratings
• Double the life
• No reduction in efficiency

58
Components Brake

• Electroid Model MRFSB-150-16
• Release mechanism will need augmentation to
  penetrate housing
• Drive shaft hub through hole can be enlarged up
  to Ø1.25 with any keyed or spline profile
• Supply voltage can be changed to 48VDC
• Release has detent which releases when power is
  supplied
• Brake material Kinel 5504
• Polyimide material
• Mfg Thermech Eng. Crop
• Per MSFC 82496

59
Components Incremental Encoder

• Stegmann model HG900E
• Reader supported by bearings in respect to disk
• Clamps directly to shaft and uses torque reaction
  bar to stabilize
• No external housing reduces bending moment load
  on drive
• Through hole size 30mm (1.18)
• Company has space experience with lower quality
  version of same model (K15) on space station

60
Back Up Components,Absolute Encoder
61
Components Absolute Encoder

• BEI Model µS14/40L Absolute Through Hole Encoder
• Ø1 through hole
• Reader supported by bearings in respect to disk
• Robust design with much space history

62
Components Motor

• Kollmorgen Model RBE-03010A-00
• Brushless design with integrated Hall effect
  sensors
• Frameless design permits for efficient packaging
  with other components such as harmonic drive,
  brake and encoders
• Company has much space and flight experience

63
Components V-Clamp

• 40º V-Clamp System to Aerospace standard AS1895
• History of accessory mounting applications
• Very high preload
• Approximately 17,000 lbf depending on clamp
  material
• Keeps joint rigid preventing stressing internal
  components
• Pilot feature ensures proper alignment
• Failsafe latch device protects against bolt
  failure
• Joint can be disassembled without removing nut
• Available in A286, Titanium, Aluminum or Inconel
• Multiple vendors including Voss Aerospace and
  local EGG Pressure Science

64
PXL Assembly and Testing
65
PXL Underwater Operations
66
Ranger Control Station
67
Ground Control Station
Video Rack
Operator Console 1
Operator Console 2
68
Control Stations Overview

• Ranger was designed from the outset to be
  controlled from the ground
• Ground Control Station Functions
• Real-time ground team control of robot
• On-ground monitoring of FCS operations
• Command, telemetry, and video recording
• Video distribution to Payload Officer
• Recording of RTSX related audio loops
• Same control station is used to control identical
  underwater robot in neutral buoyancy simulation
• Preflight training
• Inflight contingency workarounds
• Automated sequence development

69
RTSX Telemetry Distribution
Orbiter
Flight Control Station
MCC/FCR
PCC
To Rest Of NASA/ PAO
OCA Router
Payload Officer Console
Ku Ch.2 (DataCmd))
10 BaseT
10 BaseT
RS-422 (128kbps/2 Mbps)
RTSX Operator Position 1
10 BaseT
Video DeMUX and Dist.
NOTE Video Distribution is shown on a separate
diagram for clarity
Ku Ch.3 (Digital Video)
RS-422 (10 Mbps)
RTSX Operator Position 2
10 BaseT
Ethernet Hub
MUX Note No DeMUX
10 BaseT
RTSX Graphics Engine
10 BaseT
CSR
RTSX Data Archiver
T1/T3 Modem
10 BaseT
RTSX Telemetry Display
10 BaseT
To UMD
70
GCS Audio / Video Distribution
Orbiter
Flight Control Station
PCC
RTSX Telemetry Dist.
NOTE Telemetry Distribution is shown on a
separate diagram for clarity
To NASA Video Dist. (FCR, PAO, etc.)
Ku Ch.2 (Data/Cmd))
Video Distribution Amplifiers and Matrix Switche
r
NTSC Composite
RS-422 (128kbps/2 Mbps)
Video DeMUX and Decoder
Ku Ch.3 (Digital Video)
NTSC Composite
NTSC Composite
RS-422 (10 Mbps)
RTSX Operator Consoles
MUX Note No DeMUX
NTSC Composite
RTSX Video Recorders
CSR
DVIS
T1/T3 Modem
NTSC Composite
RTSX Video Display
To UMD
71
GCS Services / Interfaces

• Power
• GCS is powered from 110 VAC standard wall outlets
• Total power consumption 5200 W
• Floor Space
• Operator Consoles approx 60 ft2
• Video Rack approx 20 ft2
• Ethernet comm to spacecraft via TDRSS (equivalent
  to OCA bent-pipe to orbiter)

72
GCS Components Video Rack

• Video Rack Functions
• Decodes / demutiplexes the Ku Ch.3 downlinked
  video
• Records 5 video feeds (4 downlinked, one ground
  generated)
• Amplifies and distributes video to rest of GCS
  and Payload Officer
• Provides constant video preview of downlinked
  channels
• Records RTSX voice loops on VCRs

73
GCS Components Ops Console 1

• Operator Console 1 Functions
• Provides real-time control of Ranger
• Displays telemetry of operators choice on SGI
  display
• Interfaces with the translational and rotation
  hand controllers as primary mode of robot control
  (see Control Station Software section for more
  info)
• Interfaces with 3D advanced input device --
  Specific device TBD
• Displays video of operators choice on video
  displays

74
GCS Components Ops Console 2

• Operator Console 2 Functions
• Provides real-time control of Ranger
• Displays telemetry of operators choice on SGI
  display
• Interfaces with the translational and rotation
  hand controllers as primary mode of Robot control
  (see Control Station Software section for more
  info)
• Interfaces with 3D advanced input device
• Displays video of operators choice on video
  displays

75
Predictive and Commanded Displays
Commanded Display
Predictive Display

• Commanded displays use supervisory control
  methods and display the command sent to the
  actual system
• The display is integral in the control of the
  system
• Requires on board processing to close the loop
  about the displayed command
• Commanded simulation can be simple requiring
  little processing, processing load is placed on
  board
• Does not show dynamics, only the steady state
  solution
• Can be used to control the vehicle in real time,
  or can be used off-line to develop a script of
  actions

Predictive displays run a simulation to
forecast what the actual system will do Only
as good as the model of the simulation and its
error calibration technique Dynamic simulation
can require a large amount of processing
Simulation typically runs faster and is updated
by actual sensor data to calibrate any errors
Can show dynamics and transient motion of the
system Useful for handling force and contact
operations
76
Time Delay and Sampling Rate
Sampling Rate Across Time Delay and Display Method
The Modified Fitts Law task was used to
determine the effects on performance due to time
delay, the usage of commanded and predictive
displays, and command sampling rate Time delay
increased completion time, at 0.01 statistical
significance The effects of unmitigated time
delay caused a linear increase in completion
time Decreasing sampling rate degraded
performance, at 0.01 statistical significance for
each level of reduction A knee in performance
occurred at a sampling rate of 5 Hz, below that a
substantial effect on completion time existed
Unmitigated time delay was more affected by
reduced sampling rate, even at rates above 5 Hz
Unmitigated Time Delay Effects
77
Commanded and Predictive Displays
Time Delay and Display Method Effects

• The commanded display severely reduced the
  performance degradation with 0.01 statistical
  significance
• The commanded display reduced time delay effects
  on completion time up to 91 at 1.5-second delay
• Subjects controlled the manipulator more
  accurately with the commanded display
• Impacts were detected and compensated faster
• The predictive display also had better
  performance than time delay alone, at 0.01
  statistical significance
• The minor calibration errors caused the
  predictive displays to be about half as effective
  as the commanded display, a 0.01 statistical
  significant difference

78
Impact Comparison
Commanded Displays Reduction of Impacts

• Time delay and predictive display usage had no
  statistical significant effects on number of
  impacts
• Use of the commanded display dramatically reduced
  errors, at 0.01 significant level, even when
  compared to no time delay
• Only 3 errors were made with a commanded display
  over 4 hours of testing including 4 subjects
  testing a total of 1440 trials.
• 20 times more errors were made without a
  commanded display
• This reduction was due to subjects carefully
  positioning the commanded display to avoid an
  impact

79
Erroneous Commanded Displays
Fixed Offset Error/Varying Time Delay
Random Error/Varying Time Delay

• The peg-in-hole task was used to determine the
  usefulness of erroneous commanded displays in
  ameliorating time delay effects
• As error increased performance decreased, however
  the erroneous display still performed better than
  when no commanded display existed
• The above performance curves are statistically
  significant at the 0.01 level, except between the
  0.1 random error and the error free actual
  display only.
• The increase in completion time at 0.5 second
  time delay may be due to the small delay or to
  the commanded display occluding the actual
  display
• Number of impacts increased as the random error
  increased
• Fixed error had a significant effect that didnt
  scale with increased error

80
Manipulator Speed
Truncated Speed Limit Comparison

• The peg-in-hole task was used to compare various
  maximum allowable manipulator speeds
• 1 and 2 inches/second speeds were slower than
  other speeds with 0.1 statistical significance or
  better
• No preferred speed was found for the truncated
  speed limit
• Only 10 time reduction separated the fastest
  from the slowest completion time between 2 and 9
  inches/second truncated manipulator speeds
• A preferred speed of 7.5 inches/second was found
  when compared to all other scaled speeds at the
  0.1 significance level or better
• Subjects previous experience concentrated at 7.5
  inches/second speed, which may have influenced
  the results
• Only 5 difference separated the times between 5
  and 9 inches/second scaled speeds
• Number of impacts increased when operating at 9
  inches/second

Scaled Speed Limit Comparison
81
Summary Control Scheme Study

• The peg-in-hole task was used to reinvestigate
  the effects of time delay, commanded display
  usage, error, manipulator speed, and output
  method
• As with most previous studies, time delay was
  revisited to quantify its performance degradation
• The amelioration effect of the commanded display
  was retested as well
• One of the larger random error treatments was
  tested with zero error it was expected that the
  error would degrade performance
• The level of random error was high enough to
  create a meaningful effect without overwhelming
  the subject
• The random error was tested with all other
  treatments not just the commanded display
  treatments, this simulated a malfunction in the
  system that the subject would compensate for
• The two manipulator speeds evaluated, 3 and 6
  inches/second, showed a small difference in the
  manipulator speed study that could be enhanced by
  interaction effects
• Stereo and monoscopic displays were tested using
  the CrystalEyes LCD glasses subjects were also
  tested on a 2-D monitor
• It was believed that stereo would enhance
  performance, and that the use of CrystalEyes
  would increase simulator sickness
• All output methods had the same screen resolution
  and frame rate
• This study also focused on any interaction
  between the above listed effects

82
Time Delay and Commanded Display
Impacts Comparison
Completion Time Comparison

• Results found that differences between the
  monitor, CrystalEyes without stereo, and stereo
  CrystalEyes were negligible
• Previous results were confirmed, all completion
  time comparisons between zero time delay, 1.5
  second delay with commanded display, and 1.5
  second delay with no commanded display were
  statistically significant at the 0.01 level
• With no induced error, the commanded displays had
  statistically significant reduction of impacts

83
Effects of Error
Completion Time Comparison
Impacts Comparison

• Error within the system degraded performance
  (statistically significant at the 0.01 level)
• The effects of error were more dramatic with the
  commanded display, averaging a 60 increase in
  completion time when using the commanded display
  compared to 17 increase without a commanded
  display
• Even with the error, the command display was
  still effective relieving the effects of time
  delay
• Due to the random error algorithm, the effect on
  impacts was greater for the 3 inches/second
  manipulator speed increasing impacts by 450
• 6 inches/second manipulator speed was also
  affected with 77 increase in number of impacts

84
VIIManipulator Speed Comparison
Completion Time Comparison
Impacts Comparison

• The error and speed interaction was reaffirmed,
  error caused greater degradation with the slower
  3 inches/second manipulator speed
• With no error, less impacts occurred with the
  slower manipulator speed
• The addition of error flipped the results,
  impacts increased with the slower speed with 0.01
  statistical significance or better.
• Overall, the faster manipulator speed was
  effectively used by the subjects to lower
  completion times
• When the manipulator speed was doubled an 8
  reduction in completion time occurred with no
  error and a 15 reduction occurred with error
• When polled two subjects enjoyed the faster
  speed, two liked the slower speed, and two did
  not have a preference

85
Overall Ranking of Effects
Impacts Comparison
Completion Time Comparison

• The ranking of effects for completion time from
  most to least important was the following time
  delay, command display usage, error, manipulator
  speed, and output method
• For impacts the ranking of effects were the
  following error, commanded display usage,
  manipulator speed, time delay and output method

86
Learning Effects
Initial Learning Curve
Hand Controller Usage with Experience

• Two subjects performed four test sessions with
  the peg-in-hole task, using stereo vision, 1.5
  second delay and a commanded display with random
  error at 6 inches/second manipulator speed
• A completely naïve subject exhibited 83
  learning, while a subject with extensive robotic
  experience who had never performed the
  experimental task had 93 learning
• Hand controller usage showed that with increased
  experience, subjects spent less time waiting
• Using 93 learning as a gauge, one experiment
  needed 10 hours of training which would cut
  completion times in half, and an additional 20
  hours of experimental testing results in 8
  reduction of completion time
• This could cause a bias for test cells taken at
  the end of the experiment

87
Conclusions (1 of 2)

• Time delay increased completion time linearly
  this linear relationship occurred for both
  commanded display and unmitigated treatments
• Commanded displays without error alleviated the
  majority of time delay effects, up to a 91
  reduction
• Even with no delay the commanded display reduced
  completion time by 22
• Impacts were faster to detect and compensate for
  using the commanded display
• The commanded display facilitated more accurate
  control even when compared to no time delay, for
  the Modified Fitts Law Study
• Only 3 errors were made with a commanded display
  over 4 hours of testing including 4 subjects
  testing a total of 1440 trials.
• 20 times more errors were made without a
  commanded display
• A correlation was found between hand controller
  usage and completion time the longer the
  subjects did not move the controllers the more
  time a task took
• The move and wait strategy was evidence, at
  larger delay subjects spent more time waiting
• The commanded display enabled the subjects to
  effectively use the hand controllers during time
  delay, resulting in lower completion times
• Moderate levels of random error caused subjects
  to ignore the commanded display for fine
  maneuvering and revert to a move and wait
  strategy however, the commanded display still
  saved time during coarse movements

88
Conclusions (2 of 2)

• A commanded display was shown to be useful even
  with large amounts of error
• Evidence indicated that subjects pulsed hand
  controllers during time delay operations
• Although manipulator speed was a factor in
  lowering task completion time, no specific speed
  presented itself as superior
• The ability to move the viewpoint was more
  powerful than stereo vision at improving the
  subject depth perception
• The use of LCD shutter glasses outperformed the
  HMD tested for stereo viewing its greater
  resolution and more natural optics may outweigh
  the greater head tracking provided by the HMD for
  a given application
• Simulator sickness symptoms occurred most with
  the HMD, next the LCD shutter glasses, and least
  with the monitor

89
Future Research

• Reevaluate stereo viewing using different output
  devices newer HMDs, LCD glasses, cave
  technology, and prism screen monitors
• Concentrate on the effects of display resolution,
  field of view, and frame rate on performing a
  task with a commanded display
• Look at using a variety of ways to control the
  manipulator speeds, moving faster when accuracy
  is less important
• Investigate different methods to control the
  manipulator and the view inside the virtual
  environment head tracking, hand controllers, 6
  DOF mice, forceballs, touchscreens, and
  mechanical master arms
• Determine how to build the best commanded display
  using transparency, wireframes, flashing, and
  color to create an overlay that is helpful
  without occluding the actual display
• Compare the effectiveness of the commanded
  display against the predictive display with
  dynamic systems and contact operations
• Research should be directed to how best to use
  commanded displays for different applications,
  including using the command display with other
  advanced supervisory control methods
• Using commanded displays to control multiple
  systems
• Use the commanded display to plan, simulate, and
  alter scripts which are then run autonomously by
  the remote system
• Using a commanded display to overlay on live video

90
Ranger Spacecraft Servicing System
91
Rangers Place in Space Robotics
How the Operator Interacts with the Robot
How the Robot Interacts with the Worksite
92
Missions Enabled by Space Robotics
How the Operator Interacts with the Robot
Missions Supported by Ranger Flight
How the Robot Interacts with the Worksite
93
Engineering Arm Performing HST ECU Task

• 200 Bare bolt drive (BBD) on arm
• 055 Arm (with BBD) moves to ECU
• 051 BBD on lower left bolt
• 048 BBD turns lower left bolt 6 times
• 051 Arm moves from lower left bolt to upper left
  bolt
• 048 BBD turns upper left bolt 6 times
• 113 Arm moves from upper left bolt to upper
  right bolt
• 048 BBD turns upper right bolt 6 times
• 059 Arm moves from upper right bolt to lower
  right bolt
• 048 BBD turns lower right bolt 6 times
• 110 Arm moves from lower right bolt to tool post
• 200 Right angle drive (RAD) on arm
• 200 (Tether loop gripper on other arm)
• 057 Arm (with RAD) moves to connector drive
  mechanism
• 039 RAD on connector drive mechanism
• 028 Tether loop gripper closes on tether loop
• 020 RAD turns connector drive mechanism to lift
  ECU
• 012 Tether loop gripper opens from tether loop
• 1747 Total time

94
Ranger Application to HST SM1
Time (hrs)
1800
1500
1200
900
EVA Daily Average from SM1
600
300
000
EVA Day 1
EVA Day 2
EVA Day 3
EVA Day 4
EVA Day 5
95
Impact of Ranger-class Robot on SM3A
96
Grasp Analysis of SM-3B
Numbers refer to instances of grasp type over
five EVAs Total discrete end effector types
required 8-10
97
Results of Robot Dexterity Analysis

• Broke 63 crew-hrs of EVA activity on SM-3B into
  1860 task primitives
• 13.4 not yet categorized
• Of categorized task primitives, 95.3 are viable
  candidates for 2DOF robotic end effectors
• 71.8 1DOF tasks
• 3.2 2DOF tasks
• 20.2 tasks performed differently by robot than
  EVA (e.g., torque settings)
• 4.7 require additional dexterity
• All SM-3B robotic tasks can be performed by suite
  of 8-10 different end effectors

98
Baseline SM4 Task Allocations

• RSUs (3) 300
• Battery Modules (2) 250
• COS 310
• WFC3 255
• ASCS/CPL 330
• FGS3 335
• NOBLs (3) 150
• ASCS/STIK 155
• DSC 100
• Setup Closeout 500

99
HERCULES (Single Arm Stowed)
100
HERCULES (Dual Arm non-EVA Ops)
101
HERCULES (Dual Arm EVA Operations)
102
Approaches to Ranger Operations

• All scenarios assume Ranger dexterous
  manipulator(s) mounted on RMS throughout SM4
• Case 1 Single dexterous arm on modified MFR
• Case 1A Operates only during EVA in support of
  crew (third hand)
• Case 1B Operates only when EVA is not underway
  (conservative mode)
• Case 1C Operates with and without EVA crew
• Case 2 Dual dexterous arms on RMS
• Case 2A Mounted on MFR used only for EVA
  support
• Case 2B Mounted on MFR used only outside of EVA
• Case 2C Mounted on MFR used with and without
  EVA
• Case 2D Dedicated RMS mount both EVA crew
  free-floating
• Only Case 2C has been considered to date in any
  detail

103
Estimates of Relative Time Savings
104
HERCULES/EVA Team in SM4 Operations
105
HERCULES Proof-of-Concept Testing
106
SM4 Time Savings with Ranger Arm(s)
107
SM4R(obotic) Concept Overview
108
Maneuvering Spacecraft Bus - ICM

• Developed by Naval Research Laboratory for NASA
  ISS
• Sufficient payload on EELV for Ranger robotics,
  SM-4 servicing hardware, HST flight support
  hardware
• Sufficient maneuvering capability for extensive
  coorbital operations, followed by HST deorbit or
  boost to disposal altitude
• Currently in bonded storage at NRL

109
Dexterous Robotics - Ranger

• Developed by University of Maryland for NASA as
  low-cost flight demonstration of dexterous
  telerobotics
• Designed to be capable of using EVA interfaces
  and performing EVA tasks
• System passed through NASA Phase 0/1/2 PSRP
  safety reviews for shuttle flight
• High-fidelity qualification arms in extended
  tests at UMd SSL
• 70 of flight dexterous manipulator components in
  bonded storage at UMd

110
Servicing Option 1

• Limited to critical servicing options
• Batteries
• Rate sensor units
• Battery carrier plates, SOPE, COPE
• HST payload mass 3194 lbs
• Total ICM payload 4454 lbs
• Servicer empty mass 11,065 lb

111
Servicing Option 2

• Limited to critical servicing options
• Batteries
• Rate sensor units
• Battery carrier plates, SOPE, COPE
• HST payload mass 3194 lbs
• Total ICM payload 4454 lbs
• Servicer empty mass 11,065 lb

112
Servicing Option 3

• All SM4 ORUs and launch protective enclosures
• HST payload mass 9574 lbs
• Total ICM payload 10,834 lbs
• Servicer empty mass 17,445 lb

113
Modifications to Existing Hardware

• ICM
• Addition of TDRSS Ku-band command data links
• Mounting interfaces for robotic hardware, HST
  servicing hardware, MMS berthing ring
• Attachment to EELV payload adapter
• Ranger
• Addition of longer strut elements to provide
  needed reach for positioning leg
• Completion of flight manipulator units
• Development of required end effectors for
  servicing tasks
• Implementation of launch restraints for robot on
  ICM deck
• Development of control station for
  teleoperated/supervisory control
• HST servicing hardware
• Modification of shuttle launch restraints to ICM
  deck
• Verification of thermal environment for ORUs

114
SM4R Mission Scenario

• Launch on EELV, rendezvous and dock to HST at aft
  bulkhead MMS fittings (high level supervisory
  control)
• Perform high-priority servicing
  (batteries/gyros), other targets of opportunity
  (e.g., SM4 instrument changeouts), boost HST to
  multi-decade stable altitude
• Separate ICM and move into coorbital location to
  allow HST to perform nominal science data
  collection (no impact to HST pointing or
  stability) - ICM can be used as robotics testbed
  during this time
• ICM can redock and service multiple times if
  needed (e.g., periodic gyro replacements)
• ICM is based on design with proven flight
  duration of 6 years on-station
• At end of HST science mission, ICM redocks and
  performs deorbit/disposal boost mission

115
Launch Vehicle Considerations

• Due to size of ICM and servicing hardware, an
  EELV with a 5-meter payload fairing is required
• Delta IV Medium (5,2)
• Atlas V 501
• Also considered next larger size EELV for heavier
  mission cases
• Delta IV Medium (5,4)
• Atlas V 521

116
ICM Propellant Loads
Propellant Mass in lbs
117
Achievable Boost Altitude

• Assumptions
• 300 m/sec deltaV reserve for rendezvous and
  docking
• Remaining propellant used to raise orbit from 330
  NMi to new circular altitude, then deorbit from
  that altitude

118
Mission Assurance

• Use existing hardware to initiate comprehensive
  testing program
• Hubble SM4 EVA neutral buoyancy training hardware
• Ranger neutral buoyancy robot
• UMd Neutral Buoyancy Research Facility
• Three keys to success
• Test
• Test
• Test
• Evaluate every SM4 task in first 6-9 months and
  decide on whether or not to perform it on-orbit
• Aim for 25-30 hours of end-to-end simulation for
  every hour of on-orbit operations

119
Why SM4R?

• No other options come close to matching
  technology readiness
• ICM based on black spacecraft with flight
  heritage, currently ready to fly
• Ranger manipulators developed and tested 70 of
  dexterous manipulator flight components already
  procured
• No other options come close to matching the
  proven capabilities
• Long on-orbit endurance and high maneuvering
  capacity provide assurance of successful deorbit
  at Hubble end-of-life
• Ranger manipulators designed for EVA-equivalent
  servicing, building on 20-year heritage of HST
  robotic servicing operations
• No other options come close to matching the
  flexibility
• Interchangeable end effectors provide unlimited
  interfaces
• Ranger arm design parameters (force, speed, clean
  kinematics) unrivaled among flight-qualified
  manipulators

120
Results of a Successful SM4R Mission
Demonstration of Dexterous Robotic Capabilities
Pathfinder for Flight Testing of Advanced Robotics
Understanding of Human Factors of Complex
Telerobot Control
Dexterous Robotics for Advanced Space Science
Precursor for Low-Cost Free-Flying Servicing
Vehicles
Lead-in to Cooperative EVA/Robotic Work Sites
121
Ranger on SMV
122
For More Information
http//www.ssl.umd.edu http//robotics.ssl.umd.edu