This beta version of MFASTaaS in Portal performs initial orbit determination for multiple objects by processing uncorrelated tracks (UCTs). Order observation-to-orbit processing for one of three curated Slingshot examples.
Use MFASTaaS to evaluate the Marketplace ordering workflow and MFAST results experience without deploying or maintaining a self-hosted MFAST installation.
Note: Custom observation upload is not available in this release.
For an overview of MFAST and the wider Slingshot Processing Suite, visit the Processing Suite product page. That page describes broader MFAST and Processing Suite capabilities; the Portal Beta remains limited to the sample-only scope described below.
What can I submit?
You can choose one of these Slingshot-provided examples:
LEO Scenario
Deep-Space Scenario
Deep-Space Scenario - Small
These fixed examples and their expected outputs have been reviewed for this release. Customers cannot replace, modify, or append observations to them.
Can I upload a custom observation JSON?
Not in this release. Upload Custom JSON remains visible in the order experience with a Coming Soon label. When selected:
Upload Observation JSON, drag-and-drop, and Place Order remain disabled;
Download Observation JSON Schema remains available as a preview of the planned input format; and
the page links to contact sales if you would like to discuss custom processing or a self-hosted MFAST license.
Downloading the schema or reading the detailed MFAST-JSON Input/Output Data Model v1.8 PDF does not enable custom upload. They are provided as preparation references for the planned custom-input capability; this release still accepts only the curated Slingshot examples.
How do I place an order?
Open MFAST as a Service in Marketplace. The card displays Beta and Starts at 50 SPU.
Select Use Slingshot’s Sample JSON, then choose one of the three curated examples.
Review the sample observation details and pricing summary.
Select Place Order. This immediately creates the order; there is no additional checkout or confirmation step.
Follow the order in My Orders and download the result when processing succeeds.
Why might my order remain queued?
MFASTaaS uses a shared processing queue. An order can remain Queued while earlier jobs or currently running work use the available processing capacity. Queue wait time varies with demand, so Portal does not provide a queue-wait or completion-time estimate.
The 30-minute execution limit starts when the MFAST worker begins processing your job, not while the order is waiting in the queue.
Order statuses are:
Queued — waiting for processing capacity;
Running — MFAST is processing the selected sample observations;
Succeeded — processing completed and the result is available; or
Failed — processing timed out or another execution failure occurred.
How does SPU pricing work?
The Marketplace card displays Starts at 50 SPU. When you select Place Order, Portal temporarily holds 150 SPU:
50 SPU base charge; plus
a 100 SPU variable hold that is charged in proportion to worker runtime across the 30-minute execution limit.
The variable charge is:
100 SPU × elapsed worker runtime ÷ 30 minutes
For example, a job that uses 15 minutes of worker runtime has a 50-SPU variable charge. Its total charge is 100 SPU, and Portal returns the unused 50 SPU from the original hold.
Portal returns any unused part of the 150-SPU hold after settlement:
If the job completes in less than 30 minutes, you pay the 50-SPU base charge plus the prorated variable charge.
If the job reaches the 30-minute limit, the maximum charge is 150 SPU.
If the job fails after processing starts, the same base-plus-prorated-runtime treatment applies.
If the job fails before the MFAST worker starts, the charge is 0 SPU and Portal returns the full 150-SPU hold.
Time spent waiting in the shared queue does not consume the variable hold.
You must have at least 150 available SPU to place an order.
How does MFAST process the sample observations?
The Beta runs MFAST for one on-demand job using the selected curated sample. It performs multi-object initial orbit determination through UCT processing using the sample sensor measurements. Objects are assumed to be non-maneuvering.
MFAST uses full high-fidelity perturbed dynamics through Slingshot’s validated implicit Runge-Kutta numerical propagator, IRKProp, during complexity reduction or gating, sequential and batch state estimation, track-hypothesis scoring, and related internal processing.
The default dynamics include:
central-body harmonics using a 36x36 gravity model;
luni-solar third-body effects using ephemerides courtesy of JPL Horizons;
atmospheric drag using the NASA Jacchia 1970 model for altitudes above 90 km and NRL-MSIS for altitudes below 90 km; and
solar-radiation pressure with an umbra-penumbra shadow model.
Force-model parameters are not configurable in the Beta. The curated samples were prepared for this processing profile. Future custom observations will need to represent objects whose dynamics are consistent with these models, or sufficiently close within reasonable process-noise limits.
How were the sample observations prepared?
Successful orbit determination assumes observations were collected in short sequences representing sensor tracks, such as sensor dwells or passes, with enough cadence to establish tracks and orbits. The curated examples were prepared to exercise this workflow.
For optical data, typical guidance is:
at least three observations in each track; and
at least three tracks before the first initial orbit can be established for an object.
This guidance is informational for the current sample-only release. Customers cannot upload observations yet.
What does MFAST return?
MFAST estimates orbital states in the J2000 Earth-Centered Inertial (ECI) frame, also called the MEME-J2000 or EME2000 frame, and provides them in:
Cartesian state space; and
equinoctial orbital-element space.
The Beta returns the approved native MFAST output package for the selected sample:
output/result_orbits.jsoncontains the candidate orbits established by MFAST. Each orbit’sobs_idsfield identifies the sample observations associated with that candidate orbit. Theobject_idvalue is0because these candidate orbits have not been assigned an external catalog identity.To create valid Two-Line Element (TLE) records for the candidate orbits, MFAST assigns synthetic NORAD-style identifiers in the 80000 range within
tle_line_1andtle_line_2. These identifiers support TLE formatting only; they are not external catalog or SpaceTrack identities.output/result_uct_obs.jsoncontains the sample observations that were not associated with a candidate orbit during the MFAST run. In this article, UCT observations means observations that were not associated during the run; it does not mean that MFAST assigned those observations to an unknown catalog object.The package also includes a run summary, progress and diagnostic logs, and a README identifying the MFAST image used.
The Beta does not provide prior catalog entries to MFAST or correlate candidate orbits to an external catalog.
By default, the Beta returns quality-passing orbits with:
at least 12 associated observations;
at least four associated tracks;
at least a 0.5-day observation span; and
position uncertainty of no more than 10 km at 1-sigma.
A job can complete successfully without producing a quality-passing orbit. In that case, the order shows No qualifying orbits found and still provides the native result package.
What are the Beta limitations?
This Beta provides MFAST UCT processing only.
This release runs only LEO Scenario, Deep-Space Scenario, and Deep-Space Scenario - Small.
Customer observation upload is Coming Soon and cannot be used to place an order.
Prior catalog input, including prior orbital state vectors for known objects, is not supported.
Output associations trace only to the selected sample observations. MFAST assigns internally generated IDs to established orbits.
Maneuver detection and maneuver estimation are not included. Objects are assumed to be non-maneuvering.
Advanced force-model and processing configuration is not available.
Low-altitude low-Earth-orbit observations can produce orbital-state estimates with reduced accuracy. Review this Science guidance when interpreting sample results.
Queue duration and successful completion are not guaranteed. The 30-minute limit is an execution guard, not a completion-time promise.
A future release may add maneuver-detection and other advanced features.
Is this the self-hosted MFAST license?
No. MFASTaaS is a shared, on-demand Beta service. It is separate from the self-hosted MFAST license.
If you are interested in hosting MFAST in your own environment, use the Contact Sales option on the Marketplace offer.