Applied deep tech / Physical AI / Autonomous systems
Intelligence for the physical world.
Constela develops hardware and software that let physical machines perceive, reason, coordinate and adapt.
Live simulationX +0.00 Y +0.00
Most of the difficulty in robotics lives at the seams, where software meets hardware and one machine has to work with another. We design across those seams instead of around them, so the same foundation can serve a diagnostic instrument, a fleet of robots or an automation cell.
We are an early-stage company. The programs below are in development, each is labeled with its real status, and this site will show the work as it reaches hardware.
One foundation under every program
Constela Core is the shared technology our programs are built on. It grows only as real products need it, and it comes down to four capabilities.
- Perceive
- Read the physical world through cameras, sensors and estimation, on one synchronized clock.
- Reason
- Turn observations into events, decisions and plans that an engineer can inspect.
- Coordinate
- Share state between machines so they can divide work and cover for each other.
- Adapt
- Adjust behavior and calibration when conditions change, and recover safely when they change too far.
Safety-critical motion stays deterministic and independently protected. AI informs those systems; it does not override them.
Three programs, one company
Each program applies the same foundation to a different problem. They are at different stages, and we say which.
- InsightIntelligent diagnosticsA diagnostic system that watches a machine from the outside and reconstructs what happened when it fails.Open program
- NexusCooperative robotics and distributed autonomyResearch into robots, drones and fixed sensors that work as one system, sharing what they see and dividing the work.Open program
- FlexCellPortable, adaptive automationA robotic automation cell you can move, set down and put to work without building a fixed line around it.Open program
Research-driven, application-focused
Progress in this field comes from hard problems in real conditions. We hold ourselves to four engineering principles.
- Intelligence
- Systems that interpret physical information and support real decisions.
- Coordination
- Independent machines, sensors and devices that work as one.
- Modularity
- Hardware and software built to be reused across applications.
- Accessibility
- Lower cost, less complexity and less time to deploy. This is our aim, and we will publish evidence as we earn it.