feat(PPO_Bot): bot-relative bullets + scan staleness (STATE_DIM=57)

- Bullet state (indices 44-55): enemy-relative → bot-relative frame
  (bot needs threat vectors to itself for dodging, not to enemy)
- New index 56: scan staleness = min(ticksSinceLastScan / 30, 1.0)
  (gives policy a confidence signal for enemy data freshness)
- warm_start.py updated: 44→57 dim expansion, TARGET_DIM variable
- Tests updated for new state layout

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-08-20 14:27:28 +02:00
parent 6ad51148f4
commit 12624d3069
7 changed files with 718 additions and 16 deletions
+9 -7
View File
@@ -84,7 +84,7 @@ block testStateVectorLength:
arenaWidth: 800.0, arenaHeight: 600.0,
)
let sv = buildStateVector(bot, t)
check sv.shape == [56], "state vector has 56 elements"
check sv.shape == [57], "state vector has 57 elements"
block testStateVectorRange:
var t = initEnemyTracker()
@@ -95,7 +95,7 @@ block testStateVectorRange:
arenaWidth: 800.0, arenaHeight: 600.0,
)
let sv = buildStateVector(bot, t)
for i in 0 ..< 56:
for i in 0 ..< 57:
check sv[i] >= -2.0f32 and sv[i] <= 2.0f32,
&"sv[{i}]={sv[i]} out of [-2,2] range"
@@ -161,6 +161,8 @@ block testHistoryPaddedWhenEmpty:
# indices 44-55 (bullet slots) default to 0 when no bullets provided
for i in 44 ..< 56:
check sv[i] == 0.0f32, &"bullet slot {i} should be 0 when no bullets"
# index 56 (staleness): no contact so ticksSinceLastScan=0 → 0/30 = 0
check sv[56] == 0.0f32, "sv[56] staleness should be 0 when no contact"
block testBulletSlots:
var t = initEnemyTracker()
@@ -171,15 +173,15 @@ block testBulletSlots:
arenaWidth: 800.0, arenaHeight: 600.0,
)
# Bullet at (300,250), power 1.0 → speed = 20-3 = 17
# relX = 300-400 = -100, relY = 250-300 = -50
# dist to enemy = sqrt(100^2+50^2) ≈ 111.8, ticks ≈ 111.8/17 ≈ 6.58
# relX = 300-200 = 100, relY = 250-200 = 50 (relative to bot, not enemy)
# dist to bot = sqrt(100^2+50^2) ≈ 111.8, ticks ≈ 111.8/17 ≈ 6.58
let b = BulletData(x: 300.0, y: 250.0, power: 1.0)
let sv = buildStateVector(bot, t, 0.0, 0.0, [b], 1)
check abs(sv[44] - (-100.0/800.0).float32) < 0.001f32, "bullet relX"
check abs(sv[45] - (-50.0/600.0).float32) < 0.001f32, "bullet relY"
check abs(sv[44] - (100.0/800.0).float32) < 0.001f32, "bullet relX"
check abs(sv[45] - (50.0/600.0).float32) < 0.001f32, "bullet relY"
check abs(sv[46] - (17.0/20.0).float32) < 0.001f32, "bullet speed norm"
# second slot should be zero-padded
for i in 48 ..< 56:
for i in 48 ..< 57:
check sv[i] == 0.0f32, &"unused bullet slot {i} should be 0"
echo "All tests passed"