layup

Orbit fitting at LSST scale

Setup

Requirements

If pip install -e . fails with "File setup.py or setup.cfg not found", your pip predates PEP 660: run pip install --upgrade pip first. Note that the python3 shipped with macOS is too old; install a newer Python before creating the environment.

Before installing layup, it's a great idea to create a virtual environment with either conda or venv.

You can download the source code with:

git clone --recursive https://github.com/Smithsonian/layup.git

The --recursive flag matters: layup vendors eigen and autodiff as git submodules under include/, and the C++ build fails with a bare fatal error: 'Eigen/Dense' file not found if they are absent. If you already cloned without it, run

git submodule update --init

(assist and rebound are not submodules — they are installed as ordinary Python dependencies.)

Next, enter the layup directory and run

pip install -e .

to create an editable install of layup. If you're doing development work, you can install with

pip install -e ".[dev]"

to install all of the development packages as well.

Running the tests

Run layup bootstrap first — a large fraction of the suite is skipped without the ephemeris and reference data, so a run that has not bootstrapped will report success while validating none of the orbit fitting. Then:

pytest

Adding new submodule

Note that to get the new submodules added in an existing copy of the repo you want to run

git submodule update --init

And in subsequent clones of the repo you want to run

git clone --recursive https://github.com/Smithsonian/layup.git

Quickstart

Once layup is installed, download the ephemeris and reference data it needs (SPICE planetary kernels, the small-body kernel, MPC observatory codes, and the astrometry debiasing tables). This is a one-time download of roughly 1 GB, which expands to about 1.5 GB on disk:

layup bootstrap

Fit an orbit from the command line

layup bundles a demo dataset. Copy it into your working directory and print the matching example command with:

layup demo prepare orbitfit
layup demo howto orbitfit

prepare writes holman_data_working.csv — 4135 astrometric observations of asteroid (3666) Holman, in ADES CSV form — to the current directory, and howto prints the ready-to-run command. Fit it with:

layup orbitfit holman_data_working.csv ADES_csv -t demo_orbitfit_output

This writes the best-fit barycentric Cartesian orbit and its covariance to demo_orbitfit_output.csv. Supported input formats are MPC80col, ADES_csv, ADES_psv, ADES_xml, and ADES_hdf5.

Convert the result to another orbit representation (Cometary, Keplerian, …):

layup convert demo_orbitfit_output.csv KEP -t demo_orbit_kep

Predict future on-sky positions, with uncertainties, for an observatory:

layup predict demo_orbitfit_output.csv --days 30 --station X05 -t my_predictions

Every verb takes --help for its full set of options (engine choice, IOD method, non-gravitational parameters, parallel workers, …):

layup orbitfit --help

Control how many CPUs layup uses

--num-workers (CLI) and num_workers= (API) default to -1, meaning decide automatically: $LAYUP_NUM_WORKERS if set, otherwise 1 when layup is already running inside another worker process, otherwise the CPUs available to this process.

Set LAYUP_NUM_WORKERS when layup does not own the whole machine — running it from your own process pool, or as one of several jobs on a shared node:

export LAYUP_NUM_WORKERS=4

Otherwise each copy would size its pool to the whole machine and oversubscribe it. This is separate from OMP_NUM_THREADS and friends, which control threads within a worker rather than the number of workers.

Use the Python API

The same load → fit → convert → predict workflow is available directly from Python. See the worked-example notebook docs/notebooks/orbit_fitting_api.ipynb and the full documentation at layup.readthedocs.io.

Note that a plain pip install -e . does not install Jupyter — it is in the dev extra. To run the notebook locally, install with pip install -e ".[dev]".