Star Tracker Attitude Determination (space-systems/adcs/star-tracker)

Use when the task is star tracker attitude determination: matching measured star centroids to a star catalog, computing the angular separation between a measured centroid and its catalog star, choosing the boresight and field of view, or deciding between lost in space and tracking mode.

Domain quick reference

• A star tracker is the ADCS attitude reference with the highest pointing accuracy: it images stars, measures their centroid positions on the focal plane, and matches the pattern against an onboard star catalog to recover the instrument attitude.
• Star identification is the matching step: each measured centroid is compared against catalog stars by angular separation theta = acos(dot(u_meas, u_cat)) between unit vectors, and the best match inside the field of view is accepted.
• The boresight is the tracker optical axis, the center of the field of view. The field of view (FOV) is the angular window the sensor sees; a square FOV of half-angle fov/2 accepts catalog stars within that radius of the boresight.
• Lost in space mode identifies stars with no prior attitude by searching the full star catalog; tracking mode uses the prior attitude to predict which catalog stars should appear and matches within a small tracking radius.
• Pointing accuracy is anisotropic: cross-boresight error is set by centroid noise relative to focal length, while roll error about the boresight is larger because star positions on the focal plane barely change under roll.
• Update rate: a star tracker produces attitude at rates from a few Hz to tens of Hz; the ADCS uses the matched-star angular separation as the pointing error bound between updates.

Workflow

1. Normalize the catalog star unit vectors and the measured centroid unit vector.
2. Compute the angular separation between each measured centroid and every candidate catalog star: acos of the dot product, clamped to [-1, 1] before the acos call.
3. Accept the nearest catalog star when its separation is within the FOV half-angle; return the matched star id and the separation as the boresight error.
4. Select the mode: tracking when a prior attitude exists and the nearest match is within the tracking radius, otherwise lost in space against the full catalog.
5. Report the boresight error in arcseconds and compare the achieved pointing accuracy against the requirement.

Pitfalls

• Comparing unnormalized vectors: the dot product must use unit vectors or the acos separation is meaningless.
• Letting floating point round-off push the dot product outside [-1, 1], which makes acos return nan; clamp first.
• Accepting a match outside the FOV: the nearest catalog star is not a valid identification when its separation exceeds the FOV half-angle.
• Choosing lost in space when a valid prior attitude exists: tracking mode is cheaper and more robust.
• Reporting roll accuracy as if it equaled cross-boresight accuracy.
• Confusing the field of view (angular window, half-angle fov/2) with the boresight (its center direction).

Behavior contract (gate 3)

The star identification, boresight, and mode logic is exercised by the gate 3 contract test: scripts/test_star_tracker.py against scripts/star_tracker_logic.py (stdlib unittest, offline). Run: python3 scripts/test_star_tracker.py

Compliance

• ECSS (European Cooperation for Space Standardization) standards are freely downloadable, copyright ESA; cite the source and paraphrase. This leaf cites ECSS as reference only per standards-map.yaml; the logic here is generic star tracker geometry, not ECSS text.
• compliance: STANDARDS-REF, gated: false.