1 of 10

1
STARDUST Project CRITICAL EVENTS READINESS
REVIEW COMET P/WILD 2 ENCOUNTER Autonomous
Nucleus Tracking Shyam Bhaskaran
JPL 303-411 AM / 301-427 PM LMA MSA 1 818
354-1275 0800 AM - 0500 PM PST 20 Nov 2003
2
Nucleus Tracking Rationale

• Delivery of spacecraft using ground based
  radio/optical navigation good to about 16 km
  crosstrack and 600-1100 km (100-180 sec)
  downtrack (1 sigma) at the E - 18 hour point.
• Sufficient to maintain spacecraft safety.
• Not sufficient, however, to keep visual lock on
  comet for close approach science imaging. In
  particular
• Crosstrack uncertainty not low enough to
  determine flyby plane for the mirror to sweep as
  comet goes by.
• Downtrack uncertainty too large to determine
  mirror angles needed to maintain lock through
  closest approach.
• Improved knowledge of target relative spacecraft
  trajectory in both crosstrack and downtrack
  directions is possible using images of the comet
  as spacecraft approaches it, but 40 minute round
  trip light time prevents ground-in-the-loop
  tracking.
• Onboard closed loop tracking only option

3
Setup

• Camera FOV 3.5x3.5 deg
• Rotating mirror reflects light path from s/c X
  (mirror angle 0 deg) to s/c -Z (mirror angle
  90 deg) to s/c -X (mirror angle 180 deg),
• From angles 0 to 18 deg, the light path also goes
  through a periscope to see around the front
  Whipple shields
• Periscope incurs additional light attenuation
• Between about 6 and 16 deg, split light path also
  results in image doubling

4
Nucleus Tracking Procedure

• Onboard tracking initiated at E-30 minutes.
• Algorithm initialized with nominal values for the
  following
• Initial target-relative state (position and
  velocity), and associated covariance, at E-30
  minutes, based on results from ground-based
  navigation using E-12 hour opnav
• Camera characteristics.
• Predicts for camera pointing for all picture
  opportunities based on nominal comet-relative
  trajectory.
• Images taken at 10-30 second intervals. As each
  image is taken,
• Image is processed to locate nucleus
  center-of-figure (centroiding).
• Pixel/line values of observed centroid
  differenced with predicted center location to
  form residuals.
• At E-15 or E-10 min, all images processed with
  Kalman filter to get updated state information.
  Following this, state update is performed after
  every image is processesed.
• At E-11 min or E-6 min, latest state information
  used to determine s/c roll angle to align s/c X-Z
  plane with flyby plane
• Inertial pointing predicts are decomposed by ACS
  into s/c attitude, mirror angles and rates to
  control mirror.
• Closed-loop tracking terminated several minutes
  past encounter.

5
Nucleus Tracking Verification

• Basic algorithm successfully used for DS1
  Borrelly flyby
• Stardust specific software tested successfully at
  Annefrank encounter last year
• Primary differences with Wild 2
• 3000 km vs 300 km flyby distance
• Annefrank images started at mirror angle of 20
  deg (so no periscope images were used) vs 0 deg
  for Wild 2
• Due to fortuitous out-of-plane errors being
  small, roll was never exercised at Annefrank
• No dust environment or coma at Annefrank
• Numerous Monte Carlo simulations performed on
  unix workstation
• Simulates realistic images, ACS gyro drift
  errors, state errors
• Does not simulate attitude excursions due to dust
  impacts
• STL testing at Lockheed Martin
• Uses fully integrated software with all s/c
  subsystems
• Limited number of simulations performed

6
Annefrank Tracking
7
Tracking Imaging Sequence

• Large uncertainty in TOF value
• /- 5 min (3 sigma) if approach opnavs successful
  and dynamic correlations hold
• /- 9 min (3 sigma) assumes no improvement from
  opnavs
• 2 image sequences designed to accommodate each of
  these cases
• At 2 days prior to encounter, the decision would
  be made as to which sequence to implement
• The nominal sequence will be chosen under the
  following conditions
• Detection of Wild 2 from spacecraft shows it to
  be within reasonable tolerances from the
  predicted location
• Ephemeris fits using combined ground and opnav
  data consistent
• Ephemeris fits using different comet dynamic
  models show consistent results

8
Tracking Image Sequence (cont.)

• Nominal sequence
• E-30 m to E-5 m 1 frame/30 s
• E-5 m to E5 m 1 frame/10 s
• E5 m to E8 m 1 frame/60 s
• Filter updates started at E-10 m
• Roll maneuver at E-6 m
• Expanded sequence
• E-35 m to E-7 m 1 frame/30 s
• E-7 m to E7 m 1 frame/15 s
• E7 m to E10 m 1 frame/30 s
• Filter updates start at E-18 m
• Roll maneuver at E-11 m
• Number of frames limited by total buffer limit of
  72 frames

9
Monte Carlo Simulation Results - Nominal
Overall probability of obtaining images within
2000 km 99.5 Average of successful images
within 2000 km 47
10
Interface Files with Lockheed Martin
11
Issues/Concerns

• Need data through periscope to determine
  throughput loss for exposure settings
• Planned image through periscope in early December
• Image doubling
• Separation of images (about 20 pixels maximum)
  roughly the same as expected size of the nucleus
• Separation also about the size of spacecraft
  deadband motion
• Not expected to cause serious problems
• Currently not simulated, either in Monte Carlo
  runs or Lockheed Martin STL runs
• Plans to simulate this effect in the very near
  future
• No open ISAs
• No 1st time events