Valasek

1
VISION BASED RELATIVE NAVIGATION FOR AUTONOMOUS
PROXIMITY OPERATIONSGuidance,
Navigation, and Control Technical Committee Lunch
n LearnNASA Johnson Space Center8 January 2004
2
VisNav PROXIMITY OPERATIONS
outline of presentation

• PROXIMITY OPERATIONS (Valasek)
• VISION BASED NAVIGATION SYSTEM (Valasek)
• NUMERICAL EXAMPLE (Valasek)
• HARDWARE AND SYSTEMS (Hughes)
• LABORATORY DEMONSTRATION (Hughes)
• SUMMARY AND CONCLUSIONS (Hughes)
• QUESTIONS AND ANSWERS

3
VisNav Personnel
research assistants
Ju Young Du, Kiran Gunnam, Roberto Alonso,
Changwa Cho
4
VisNav Personnel
senior researchers and partners

• TEXAS AM UNIVERSITY
• Dr. John L. Junkins
• Dr. John Valasek
• STAR VISION TECHNOLOGIES INC.
• Mr. Michael Jacox
• Dr. Declan Hughes
• Mr. Brian Wood
• SARGENT FLETCHER INC.
• AIR FORCE RESEARCH LABORATORY
• Munitions Directorate, Eglin AFB
• BOEING
• St. Louis

5
AUTO. PROXIMITY OPERATIONS
considerations

• Ideal Scenario
• Quickly arrive at a space object
• Perform maneuvers without
• months of planning
• intersatellite communication
• Utilize small mission operations center
• Desired Functionality
• Inspection
• Anomaly assessment
• Formation flying
• On-orbit servicing
• Requirements
• Three-axis stable
• Autonomously detect location and maintain
  attitude

6
AUTO. PROXIMITY OPERATIONS
challenges

• Sensing relative position and velocity when in
  proximity to another object
• Non-cooperative objects
• Low power consumption
• Advanced autonomous event planning
• Forward thinking resource manager
• Only two or three people should be
• required to plan and monitor the S/C

7
RELATIVE NAVIGATION
approaches

• Radar corridors
• Ground tracking updates
• Scanning LIDAR
• GPS Local Positioning System (LPS)
• Optical-based navigation systems offer promising
  alternative
• Negligible diffraction, high bandwidth, no
  droputs
• Multipath reflections minimized by restricted
  field of view
• Optimal Signal-Noise (S/N) ratio
• closed-loop control of beacon intensity
• System selects beacons from redundant set
• robustness and flexibility

8
RELATIVE NAVIGATION
pattern recognition

• Pattern recognition and visual servoing using a
  CCD camera to provide the line of sight vector
  for end game docking maneuver
• Reduced effectiveness in poor optical conditions
  (sunlight, etc.)
• Depth of Field
• camera must be able to focus over very long and
  very short distances
• Accuracy
• determination of 3D position coordinates from 2D
  images
• Requires very high camera resolution (appx. 106
  pixel data on each frame)
• Reliable pattern recognition
• greater than 85 reliability is difficult to
  achieve, even in a perfect laboratory setting
• Requires high computational speed and data
  bandwidth
• current data bandwidth is only 40 Hz with
  intermediate resolution

9
RELATIVE NAVIGATION
VisNav cooperative vision

• Optical sensor with active structured beacon
  lights that provides an accurate, high speed
  6-DOF navigation solution for the mid to end game
  docking maneuver.
• Update rate of 100 Hz and high precision under
  optimum conditions.
• Reduced risk
• Feasible at current level of optical sensing
  technology
• Concept validated with hardware in laboratory
  experiments

10
VisNav SENSOR
position sensing diode

• Activated by energy from light sources
• Generates electrical current in 4 directions
• Current imbalances are linearly proportional to
  location of image centroid

11
VisNav SYSTEM
positions and attitudes

• Ideal pin hole camera model
• colinearity equations
• Line-of-sight vector observations

Sensor on Vehicle A
IU
(yo, zo)
IR
z
Pi
IL
y
x
ISC
ID
12
DIFFERENTIAL CORRECTION
Gaussian least squares

• States
• Measurements (n ? 4)
• Measurement Sensitivity Matrix
• Measurement Residuals

13
EXTENDED KALMAN FILTER
zero acceleration

• Linearize about estimated (reference) states
• Zero acceleration model (or IMU)
• States
• Nonlinear system dynamics
• Sensitivity Matrix
• State transition matrix

14
VisNav HARDWARE
6-DOF algorithm

• 800 Hz beacon switching, 100 Hz 6-DOF update rate
• Accuracies
• 1cm/0.25 deg at 30m
• 1mm/0.05 deg at 0.5m
• Modified Rodriguez Parameters
• Good convergence
• Options
• Kalman filtering of sensor data
• Combined model/Kalman filter.
• Beacon selection criteria options, computed in
  real-time
• 6-DOF data covariance matrix condition number
• Apparent beacon selection width and depth of field

15
VisNav APPLICATION
precision landing
16
VisNav APPLICATION
autonomous aerial refueling
17
VisNav APPLICATION
autonomous aerial refueling
18
VisNav APPLICATION
data glove
19
VisNav APPLICATION
S/C docking
20
NUMERICAL EXAMPLE
21
NUMERICAL EXAMPLE
Position and Attitude errors of GLSDC
22
NUMERICAL EXAMPLE
Position and Attitude errors of Extended Kalman
Filter
23
VisNav HARDWARE
psd sensor construction

• Optical filter to block visible light.
• Wide angle lens focuses wide field of view onto
  PSD
• Approx. 3 x 3 x 3

24
VisNav HARDWARE
psd sensor

• Wide angle lens focuses wide field of view onto
  PSD
• Approximately 3 x 3 x 3

25
VisNav HARDWARE
optical filter

• Infrared LED
• close to maximum PSD response
• Thermal noise dominates at low illumination
• Shot noise proportional to sqrt(rms PSD current)
  sunlight
• large current and shot noise dominates

26
VisNav HARDWARE
micro computer

• Small computer
• micrporocessor TI TMS320VC33 DSP _at_ 60 MHz
• IMbyte SRAM
• 0.5MByte Flash Eprom
• 120 MFLOPS, 1W
• Circuit outline 2.3 x 3.3
• Analog interface circuit
• stacks on top

27
VisNav HARDWARE
active beacon

• Control signal carrier at 40 KHz
• Largest beacon is 218 LED design
• Light Shaping Defuser (LSD) positioned in front
  of LED
• Red or IR optical filter
• protects plastic LSD and LEDs from sunlight
• 1W optical
• 10W electrical

28
VisNav HARDWARE
active beacons

• Three Beacon Sizes
• Largest Beacon 218 LED design.
• Stacked board design, V-gtI circuits behind
• A few red LEDs are used for visual check

29
VisNav HARDWARE
active beacons

• Aluminum fabricated boxes
• Glass front plate (colored glass can also be
  used)
• Light Shaping Diffuser (LSD) lens.
• One push-pull connector

30
VisNav HARDWARE
new LEDs

• gt Watt emitted energy.
• Wide Variety of wavelengths, including 880nm.
• Tailor radiation pattern by cutting flat surface.

31
VisNav HARDWARE
calibration and test rig

• Yaw-Pitch axes actuator1 for sensor calibration
• X-Y-Z-Yaw-Pitch-Roll actuator2 to test complete
  system accuracy
• Beacons placed on optical table

32
VisNav HARDWARE
calibration

• Divide by beacon intensity
• Measure at many yaw/pitch data points
• Reduce data set by calculating normalized voltages

33
VisNav HARDWARE
calibration

• (L-R)/(LR)
• intensity effect removed
• Measure at many yaw/pitch data points
• Invert surface and fit

34
VisNav SYSTEM
application

• First VISNAV system application.
• Beacons placed in Nasa JSC NSTL room on walls and
  on a movable frame.
• Frame may be moved outside used for docking
  simulations.
• VISNAV will calibrate differential GPS sensor
  system.

35
VisNav SYSTEM
docking simulations

• Beacon frame contains 8x 60 LED beacons, and 8x
  12 LED beacons.
• Grey box beacon controller.
• Sensor uses large beacons when in field of view.
• As sensor approaches lower right hand corner it
  switches to smaller beacons that remain in field
  of view.

36
VisNav SYSTEM
docking simulations

• VISNAV 3.2 Schematic

37
VisNav SYSTEM
options

• Temperature sensing and compensation.
• Phase lock sensor to incoming signal gt better
  accuracy.
• Daisy chain serial beacon control cable gt less
  wiring.
• Digital transmission gt no noise pickup.
• Wireless beacon controller to beacon connections,
  or add beacon controller to each beacon gt no
  separate beacon controller, no control cables.

38
VisNav SYSTEM
wireless IR link

• IR wireless sensor to beacon controller link,
  115.2 Kbaud.
• Crystal locked FSK.
• Low latency and latency variation.
• PCB 1.8 x 1.4 approx.

39
VisNav SYSTEM
options

• Self-test, calibration LEDs in sensor.
• Accelerometers
• Wider bandwidth/less noise.
• Smaller beacons, circular shape.
• Ultrasonic version
• 6DOF relative navigation underwater. Relate to
  R/C GPS boat(s) on surface.
• 120MFLOP TI TMS320VC33 DSP, 150MFLOP version, or
  1GFLOP TI TMS320C6701 DSP card still 2.3 x
  3.3.
• Addition of LIDAR
• More accurate estimates useful at longer ranges.
• Designing passive beacon version

40
VisNav SYSTEM
smaller DSP
Small Computer uP TI TMS320C5509 DSP _at_ 200
MHz 8 Mbyte SRAM, 0.5MByte Flash Eprom, 400
MIPS, 0.5W Circuit Outline 30mm x
36mm Firewire and many other interfaces.
41
FUTURE DIRECTION
stereo vision geometry

• Determine position of UAV using LOS vector from
  two different cameras with known position and
  attitude
• States
• Measurements

42
CONCLUSIONS

• Accurate 100 Hz update rate 6DOF data possible
  with small low power sensor/beacons to 100m.
• Beacon signal modulation and optical filtering
• Excellent ambient light rejection.
• Realtime beacon selection/intensity control
• Minimize power requirements.
• Pattern recognition problem effectively
  eliminated.
• Very wide field of view, no moving parts.
• Distributed beacons
• Very large operating space, redundancy.
• Facing sun operation estimated appears feasible
  yet to be demonstrated.