Mission autonomy under human authority.
We are building the runtime for autonomous space missions: software that lets spacecraft, satellites, and planetary rovers plan, adapt, and act within limits that humans set and machines cannot cross.
A missing systems layer, not a missing model.
Space missions are gaining autonomous components faster than they are gaining ways to govern them.
A modern mission might carry an autonomous navigator, a science-targeting algorithm, a fault-management system, and a learned perception model, each built by different teams, each safe on its own terms, and no common layer that understands the mission's intent, its constraints, and its current state well enough to decide what may execute next. The missing piece is not a smarter model. It is a trust boundary.
Named parts. Real mechanisms.
The Barycenter runtime is that layer. It compiles authorized human intent into formal goals and prohibitions. It maintains a probabilistic belief state of the mission, tagged with the provenance of every input. It reasons jointly over science value, safety, power, thermal state, communication windows, and time. And nothing executes without a permit from an independent assurance kernel, with every decision recorded for deterministic replay.
Learned components have a place in this architecture: bounded, contracted, and monitored. They are never the trust boundary.
Intent compilation
Authorized operator intent becomes formal goals, prohibitions, and priorities the runtime can reason over, not English strings, not model prompts.
Mission belief state
A probabilistic representation of what the vehicle believes about itself and its environment, with the provenance of every input preserved.
Assurance kernel
An independent mechanism that issues execution permits. A permit is granted when goals, constraints, and safety envelopes are simultaneously satisfied; otherwise nothing runs.
Deterministic replay
Every decision, and the state it was made in, is captured such that an accident board can replay it exactly. Assurance is a property of the record, not a promise.
Engineering positions, not marketing.
Our positions are published as research notes, grounded in NASA technology shortfalls, three decades of JPL flight heritage, and agency roadmaps from four continents.
If your mission needs autonomy it can certify, we would like to hear from you.
We are working with a small number of design partners on reference missions. Partners get early access to the reference implementation and direct influence on the roadmap. In return we ask for real mission context and honest feedback.