COSE · AstroBotany · in-silico tropisms
Same simulation as v1, but the root-branching rule is now calibrated against
real root-image data: 24 RootNav/SmartRoot-traced cotton root scans
(astroroot's samples/tictoc)
and a mechanistic auxin-transport gravitropism curve
(virtual-root). Branching now fires from a
stochastic interval (not a bend heuristic), lateral angles are drawn from the real annotated
distribution and alternate sides, gravitropic gain follows the real saturating dose-response
curve, and roots can bounce off or grow over each other on contact. See
data/calibration.json and
MANUSCRIPT.md for the full derivation.
Micropyle = spread of initial root-exit angles across the seeds. Each seed's coat now visually rotates to face its own micropyle direction — seeds aren't perfectly round, so the coat shape should reflect which way the radicle will emerge.
Gravitropic gain now follows the real saturating asymmetry curve from the virtual-root auxin model (µG→Moon→Mars→Earth→2g = 0/10/21/36/41% lower-side asymmetry), not a linear ramp — response is much weaker below 1g than v1 implied. Directional noise also increases as gravity→0 (OSD-121: root-angle dispersion 51.2° flight vs 45.3° ground).
Colour is visual only — this model responds to light intensity & direction, not wavelength.
When two roots touch: 0% = always grow over/through each other (v1 behaviour); 100% = always deflect off each other; in between (e.g. 40% ≈ "4:6", 30% ≈ "3:7") some contacts pass through and some bounce, matching the mixed behaviour real crowded roots show.
Adds a burst of extra directional variance right at the moment two roots touch, on top of whatever the bounce/grow-over outcome was — models the unpredictable local response real roots show to mechanical/chemical contact, beyond simple deflection.