Degraded sensing
Poor light, weather, occlusion, and sensors that partially fail.
Method · Simulation & Test
Simulation
and test.
Most of our progress happens in software. Testing an idea in simulation is faster, safer and repeatable in a way that field testing never is — so we find problems on screen, and only move to hardware once the software holds up.
The Method
Prove it in software first.
Autonomy is developed by iteration, and iteration speed decides how good a system eventually becomes. In the field, a single test can take a day to set up and produce one noisy result. In simulation, the same scenario can be run thousands of times before lunch, with every variable controlled.
That does not make simulation a substitute for reality — it makes it the filter. By the time something reaches hardware, the obvious failures are gone, and the tests that remain are the ones that genuinely need the physical world.
The Loop
Test. Rewind.
Assess. Improve.
Our name describes the method. Every run — simulated or physical — is recorded, replayed and assessed honestly. What went right, what went wrong, and why.
- SimulateRun the scenario in software, as many times and as many ways as needed.
- Test on hardwareMove to real devices once the behaviour holds up in software.
- RecordCapture everything, so a run can be reconstructed exactly rather than remembered.
- Assess honestlyReview what actually happened without flattering the result, then feed it back in.
DevelopmentTest · Record · Review
Primary environmentSimulation
HardwareOnce software is sound
Every runRecorded and reviewable
First mission testPlanned December 2026
StageResearch & development
What We Test For
The cases that
actually matter.
Anything can be made to work in the nominal case. We spend our time on everything else.
Lost communications
What the platform does when the link thins, drops, or returns unexpectedly.
Unexpected world
Situations the system was not explicitly designed for, to see how it fails.
Edge of the envelope
Behaviour right at the boundary of the limits an operator has set.
Repeatability
Whether the same input reliably produces the same behaviour.
Hand-off
Transferring control between autonomy and operator without confusion.
Our systems have not been fielded. Everything to date has lived in development and simulation; our first end-to-end real-world mission test is planned for December 2026. We would rather say that plainly than imply more.
Go Deeper
Rigour is
the product.
If disciplined testing is how you like to work, you'll fit here.