From a catalog to a close approach, in four steps
A schematic. A few hundred made-up dots stand in for the 15,968 objects, the links and their weights are for show, and nothing is to scale.
Close approaches are ranked by collision probability into 5 risk tiers. One day of them is screened in 19 s.
- A pale dot is an object in the catalog.
- Coral is a possible close approach: a link between two candidates, the two ringed objects being followed, and the circle round the gap between them. A heavier link is a higher tier.
The problem and the idea
Commercial satellite conjunction tools are expensive and closed. ORCA is an open pipeline from raw orbit data (TLEs) to the close approaches that matter this week.
What I built
- Screens a 15,968-object satellite catalog for one day of close approaches in 19 s
- KD-tree candidate search within 100 km, then parallel refinement of each approach
- Ranks close approaches by collision probability into 5 risk tiers
- FastAPI + React/three.js, Docker on Oracle Cloud with CI deploys
The result
- 15,968Objects screened
- 19 sFor one day of close approaches
- 5Risk tiers
Narrowing the search
A flat, made-up picture of the idea: 64 invented dots and an invented radius. The real search runs over the 15,968-object catalog.
For one object, only the boxes its search radius touches are opened. In ORCA that radius is 100 km, and what it finds goes on to refinement.
In this picture: 5 of 16 boxes opened, 19 of the other 63 dots measured, 10 inside the radius.
- A pale dot is an object. The white one with a coral edge is the object being searched round.
- Coral is the search: the dashed circle is the radius, the tinted boxes are the ones opened, and the coral dots are the candidates found.
Stack
- Python
- FastAPI
- React
- three.js
- Docker
- Oracle Cloud