feat(*): Replace the test terrain arena with an enclosed standard-size space-platform arena (shared ArenaBoundary floor/walls/ceiling scene, starfield sky, ball CCD) and derive TrainingMode placement bounds from it, dropping the out-of-bounds reward guard

This commit is contained in:
Josh Creek
2026-07-18 20:18:03 +01:00
parent 85f96eb15e
commit 379ef9910e
15 changed files with 181 additions and 6492 deletions
+3 -2
View File
@@ -11,8 +11,9 @@ extends RefCounted
# The same rotation must be inverted when interpreting actions (see canon —
# it is its own inverse).
# Normalization scales. Arena bounds: goals at z ≈ ±15.56, ship spawns at
# z = ±12; positions are soft-normalized to roughly [-1, 1].
# Normalization scales. Standard arena volume (see ArenaBoundary): x ±12,
# z ±18, height 12, goals at z ±17; positions are soft-normalized to roughly
# [-1, 1]. Do not retune without retraining every model in Game/bots/.
const POSITION_SCALE := Vector3(20.0, 10.0, 20.0)
const BALL_SPEED_SCALE := 30.0
const GOAL_DISTANCE_SCALE := 40.0