Capture the site
Photos and scans of the customer's floor, taken during a normal site visit.
SEE WHERE YOUR POLICY FAILS
General Deployment is the simulation layer for testing and evaluating robot policies. We minimize the real2sim gap for you to build trust between the industries using your automation.
Request a demoSeed 1337 · Loading scene
How it works
Pipeline · 4 stepsWe help you with deployment end to end. Everything you need from hardware to post-training to testing.
Reconstruct the twin
MJCF assets, segmented objects, a deployment-realistic layout.
OUT/ site.mjcf
Sweep × run policy
Lighting, clutter, pose, occlusion.. the policy runs against each condition.
Ranked failure map
Where the policy breaks, ordered by what matters on your floor.
Condition sweep
Seed 1337 · N=20The same scene, infinite varied conditions. Situations you can't test in the physical world.
C-001PassNOMINAL · LIGHT 460 lx
C-002PassLIGHT 690 lx
C-003FailLIGHT 85 lx
C-004PassCLUTTER 4/8
C-005FailOCCL 55 %
C-006PassLIGHT 800 lx
C-007FailCLUTTER 7/8 · LIGHT 180 lx
C-008PassPOSE Δ6 cm
C-009PassOCCL 15 %
C-010PassMIXED · OCCL 20 %
C-011FailOCCL 70 %
C-012PassLIGHT 280 lx
C-013PassCLUTTER 6/8
C-014FailPOSE Δ9 cm · CLUTTER 6/8
C-015PassLIGHT 690 lx · CLUTTER 4/8
C-016PassPOSE Δ7 cm
C-017PassOCCL 25 %
C-018FailOCCL 45 % · LIGHT 130 lx
C-019PassOCCL 30 %
C-020PassMIXED · CLUTTER 5/8
14 pass · 6 fail
Aggregate metrics
Custome report analysis for your biggest deployment questions.
See your site as a digital twin.
Independent, third-party evaluation of a learned policy, in your exact environment.