What a test report tells you
Selected cabinet-shop checks, their results and limits, and the next physical question to investigate.
Read the sample report
Notes
Technical notes.
Methods, measurements and test records from our projects.
Measuring a field into the simulator
Our rover’s simulated field was built before anyone measured the real one — rows drawn at 3.05 m against a measured 2.30 m, a lane a third wider than the machine actually gets. The note covers how the real field was measured out of the rover’s own recordings and written back as an Isaac Sim stage: row spacing taken inside a single LiDAR sweep so odometry drift cannot touch it, IMU mount calibration from the direction of world-up, and how scene alignment and collision checks complement the render.
It also covers a measurement trap worth knowing about: two Gaussian-splat training runs scored a PSNR as good as a working reconstruction while producing a featureless blob, so our quality gate now measures sharpness against the real frames as well.
Read the noteRow spacing, measured2.30 m median · 156 row pairs
The hand-built field used3.05 m — 33% too wide
SurveySix recordings, four sessions
Observed sharpness ratio3.7% of real-frame detail
Next step
Working on something similar?
Book a free twenty-minute call to discuss the job. We confirm whether we can quote from your documentation or need a paid scoping visit.