Proposal: FGOC (Focal-Geometry and Curvature) Module for Early ISO Detection in LSST Prompt Processing
Hi Rubin team,
I have been studying short-arc geometric behavior with the goal of enabling earlier ISO / non-Keplerian
detection in LSST. Based on this work, I prepared a fully engineering-focused Technical Note (RTN-2025-01)
that proposes a small, self-contained module called FGOC (Focal-Geometry and Curvature classifier).
My intent is to contribute something concrete and immediately testable to LSST Prompt Processing:
a lightweight classifier that runs directly under DIASource, produces deterministic geometric–curvature
features, and can support MOPS Pre-Linker in prioritizing dynamically unusual short arcs — all without
orbit fitting and without modifying any existing AP/MOPS logic.
The attached PDF is written specifically for AP/DM engineers. It contains:
- a complete description of the module’s internal logic,
- computational cost (< 1 ms per short arc),
- API and I/O interface,
- schema extensions,
- flowcharts showing the insertion point in AP,
- and a proposed evaluation / benchmarking plan.
Summary of the FGOC module
FGOC computes four deterministic outputs from 2–5 point short arcs:
-
fgoc_flag— Boolean early anomaly indicator -
fgoc_score— 0–1 combined geometric–curvature score -
focal_axis— estimated trajectory axis -
curvature_sign— +1 or –1
Key engineering properties
- Runtime < 1 ms per short arc
- Non-invasive and fully reversible
- Zero changes to existing AP/MOPS structures
- Uses only RA/DEC/MJD + optional metadata
Feedback I would appreciate from AP / Prompt Processing / DIASource / MOPS teams
- Whether FGOC could be evaluated in a dry-run mode within Prompt Processing.
- Whether this module might be of interest for experimental integration or testing.
- Recommendations on the appropriate DM/AP/MOPS contacts for next steps.
Thank you very much for your time and consideration.
The full engineering specification is attached below.
Best regards,
Lâu Thiat-uí
Rubin Observatory Technical Note (RTN).pdf (157.8 KB)