Historical page (2026-11-07). The full-path study uses the V1 flow model and the V5-era FLOWER tube. It was replaced by V7 and FLOWER D3. See the current robot, V7 and FLOWER D3.

Full path: floor, intake, deck, windmill, flywheel, then the HIVE or the FLOWER

One chain, end to end, in simulation. The ball is shot by the flow model (MuJoCo, plan setting: roller bar 16 mm lower, pod float 60 deg, left upright moved 16 mm) and the state it leaves the robot with (position, velocity, spin) is handed to a drag flight into the HIVE, or to the 3-D tube model and drop model of the FLOWER page. The numbers are a product of model stages with assumed contact parameters. They are not measured rates.

Model speed uncalibrated: 3.3 m/s. Scenarios at 5.5 / 6.5 / 7.5 m/s keep each run's direction, spin and release point from the flow model and only scale the speed. The raw-model result is shown first everywhere, the scenarios next to it.
Ball Target
scenarios scale the speed at release; the chain before the release is always the raw model
robot, x = 0 section of the flow model CAD (ball path projected)windmill platesGecko wheels, deck feederHogback flywheelFLOWER tube

This run

One number per ball and target

Exit state the flow model actually produces, against the assumption

Launcher setting ladder (the only speed control the flow model has)

HIVE hit rate by distance and bearing

Ball

FLOWER mouth mismatch: the funnel is built for a 53 deg shot, the model shoots at about 70 deg

Frame check: is (105, 72) the flywheel?

NECTAR sideways spread

FLOWER delivery rate by wall material

End to end: intake x feed x target

What is and is not shown, and what is uncertain