ORBITRISK

Research overview

OrbitRisk is a reproducible public-data application for satellite close-approach screening. It parses an attributed CelesTrak Stations TLE snapshot, propagates compatible objects with SGP4, and reports geometric conjunction candidates.

The release is a scientific-computing demonstration, not an operational conjunction-assessment system. Real public-TLE outputs never include a collision probability because covariance is unavailable.

Problem definition

A useful research screen must distinguish changing relative geometry from uncertainty. The practical output is a time of closest approach, miss distance, and relative velocity, with explicit data epoch, source, frame, and uncertainty limits.

Data sources and preprocessing

The catalog is CelesTrak GP Stations data obtained without credentials. The original three-line TLE records are preserved with a retrieval timestamp and SHA-256 checksum. Metadata retains NORAD ID, object name, TLE epoch, inclination, eccentricity, and mean motion.

Orbital propagation

Compatible elements are propagated with the SGP4 implementation of the standard NORAD model. The returned states are TEME inertial position and velocity in kilometers and km/s. OrbitRisk does not treat raw SGP4 outputs as Earth-fixed coordinates.

Conjunction screening

Object pairs are sampled over a bounded window on a 10-minute grid. A plausible local minimum is refined with bounded scalar minimization of relative Euclidean distance. The retained event includes primary and secondary objects, refined TCA, miss distance, relative velocity, threshold, source, and uncertainty status.

Uncertainty modeling

The probability laboratory is a separate controlled 2-D encounter-plane experiment. A stated isotropic Gaussian uncertainty, hard-body radius, seeded Monte Carlo sample, and analytical noncentral chi-square calculation make its assumptions inspectable. Its probability is never associated with a real TLE conjunction.

Experimental design and validation

The shipped tests confirm that the preserved catalog parses, yields finite SGP4 state vectors, and returns deterministic seeded Monte Carlo output. The API also bounds user-controllable screen windows, thresholds, and sample counts to avoid unrestricted work.

Results

The live application exposes the current result of its CelesTrak snapshot screen rather than a baked-in statistic. It also exposes a reproducible synthetic probability comparison. Results may change when the public source is refreshed, so the snapshot epoch and checksum are treated as part of every result.

Limitations

TLE state quality, element age, atmospheric modeling, unmodeled maneuvers, and missing covariance prevent operational interpretation. Public screening must not substitute for conjunction data messages, mission-specific hard-body modeling, independent screening, or authoritative state estimates.

Conclusions

OrbitRisk demonstrates an auditable path from public orbital elements to a bounded geometric screen and an explicitly synthetic uncertainty experiment. The distinction between those modes is the central scientific constraint of the project.

Reproducibility

The public repository includes the source snapshot, parser, propagation and screening modules, synthetic experiment, tests, API, standalone frontend, PDF source, and deployment instructions. Refreshing a snapshot should be recorded before comparing any rerun.