Add tools/blender/gen_nebula.py (previously nebula_decoration.blend had
no generator script, unlike ship/ball) to rebuild the station with
inset/extrude panel-line greeble and three separate emissive window
strips, and give debris its own rockier, damage-scarred materials
instead of cloning the station's hull material.
Ship gets a greebled hull, tapered nose, swept canopy, twin engine nacelles,
and tail fin (built via the vendored Blender MCP, generator committed at
tools/blender/gen_ship.py) in place of the 5 flat primitives. The ball is
fully remodeled as a smooth round sphere with a crossed emissive accent
pattern (gen_ball.py), replacing the old flat-shaded gold_ball rather than
just tweaking its material.
Node names (Nose/TailFin) are preserved for Ship._apply_team_color(), and
the RigidBody3D/CollisionShape3D physics on both ship.tscn and ball.tscn are
untouched so RL-trained bots and flight feel stay valid. Each part's mesh is
extracted to a standalone .res (tools/blender/extract_meshes.gd) rather than
referenced via glb::ArrayMesh_xxx sub-paths, which don't reliably resolve
across scene files and were silently rendering both models invisible.
Adds sdfgi_enabled and a ReflectionProbe to all three arenas' Environment
setup so surfaces get real bounce lighting/reflections instead of flat
ambient. Converts arena_boundary.tscn's shared glass field material from
unshaded to a shaded, rim-lit material so it reads as glass (dim face-on,
highlighted at grazing angles) rather than a flat tinted overlay.
Every bright star in sky_nebula.png reused the exact same diffraction-
spike stamp, just relocated. Commit the generator (recovered from an
ephemeral scratchpad) to tools/textures/gen_nebula_sky.py, randomize
each hero star's rotation, arm count, spike length, and brightness,
and regenerate the texture.
Enable glow/bloom, color adjustments, and MSAA+FXAA across all three
arenas so emissive ship/station accents actually bleed light and edges
read cleanly. Expose the player-facing knobs (anti-aliasing mode, glow
intensity, brightness) through a new Settings screen off the main menu,
persisted via a VideoSettings autoload that scales each arena's own
tuned Environment values on load rather than overwriting them.
Introduce arena_02 (Nebula) and arena_03 (Asteroid Field) alongside
arena_01, listed in a new ArenaRegistry (scripts/arena_registry.gd) as
the single source of truth for available arenas. Free Play lets the
player pick an arena from the main menu; Match/Spectate each pick one
at random per session; Training keeps its own fixed arena_01.
Nebula gets an RL-style visual treatment: a near-invisible glass
boundary material shared by all arenas, a baked equirect nebula sky
texture, a drifting-dust particle system, and a Blender-modeled
station/debris/planet decoration set. GameMode gains a
_get_arena_scene_path() hook so modes can instantiate their arena in
code instead of hardcoding it in the scene.
Replace the raw checkpoint dropdown with curated Easy/Medium/Hard presets
that drive GameSettings' bot model/reaction_ticks/action_noise overrides.
Move raw-checkpoint testing and Spectate mode into a dev-only section
hidden via OS.is_debug_build() so they disappear from release exports.
Generation 2's single "unmask" stage (flip vertical/pitch-roll locomotion
from grounded-only to full 3D in one step) failed 3 independent 240M-step
attempts, landing at a stable 32% / 28% / 31% win rate vs curric-s5-aggression
each time -- not noise, and not fixable by more training time (attempts 2-3
each continued the same checkpoint lineage for another full 240M steps with
zero improvement). Every attempt shows train/std collapsing from ~0.30 to
~0.13-0.15 within the first ~10% of steps and never recovering: the policy
locks the newly-opened axes back down before ever meaningfully exploring
them.
Replaces the boolean allow_vertical/allow_pitch_roll mask on ShipAIController
with float vertical_ramp/pitch_roll_ramp multipliers (0.0-1.0), scaling axis
effect in set_action() instead of gating it outright -- the action space
never changes shape, so checkpoints stay resumable across ramp values. The
single unmask stage in curriculum.py becomes 4: three ungated warmup stages
(25%/50%/75% authority, airborne_penalty ramping in step) that train,
checkpoint, and always advance with no eval gate, then the measured stage at
full authority -- same reference, opponent mode, and 240M budget as the 3
failed attempts, for a direct comparison. Adds a "gated" flag/branch to
main()'s loop for the ungated stages.
This is generation 3 of the curriculum; generation 2's state is archived to
curriculum_state_gen2.json (mirroring the earlier gen1 -> gen2 archival) and
curriculum_state.json resets fresh, since its stage 0 no longer means what it
used to. See TRAINING.md's "Generation 3" section for the full postmortem,
stage table, and the open question about whether scaling action effect in
Godot (which PPO's own entropy/exploration math never sees) actually
addresses the collapse.
Generation 2's first two real stage-1 attempts both independently restarted
from curric-s5-aggression (reset_retry_checkpoint) with identical flags and
landed at 32% and 27% win rate vs the reference -- a real regression either
way, but too much spread between "identical" runs for repeat fresh restarts
to be a controlled test of anything. The first attempt's own trajectory
(ep_rew_mean climbing from -10.86 toward ~0 by the 240M-step cutoff,
briefly touching positive) looked closer to convergence than the second's,
so retries now continue that attempt's own checkpoint for another full
timesteps budget instead of resetting to foundation again.
Drops retry1 and retry2 (checkpoints, logs, exported bots, eval_history
entries) -- retry2 never trained meaningfully before crashing on the
GoalRateCallback bug just fixed, and retry1 was the inferior of the two
real samples. curriculum_state.json rewinds to attempt 1, in_progress, so
the next run resumes 20260726-1904-curric-s1-unmask/final.zip directly.
Stage 5 (aggression) passed (41-47 vs grounded curric-s2-defend, within
the lenient gate but not yet a clear win). Rather than keep the locomotion
mask on indefinitely, stage 6 reopens full 3D controls on top of the
aggression retune and pairs it with a new dense airborne_penalty (scaled
by height above the floor) so the policy learns to prefer staying grounded
through incentives instead of a hard mask — same regime shift that
regressed stage 3, but this time with a mitigation and ~12x the training
time (~240M timesteps / ~24h vs ~20M / ~2h) to actually re-converge
instead of stalling mid-shift.
airborne_penalty follows the existing SHIP_AI_OVERRIDES pattern: default
0 (off) on ship_ai_controller.gd, exposed via train.py's new
--airborne-penalty flag, added to training_mode.gd's allow-list. Also adds
a per-stage timesteps override in curriculum.py (STAGES[n]["timesteps"])
since this is the first stage to need a different budget than the rest.
GameMode.reset_ball()/reset_ships() teleported to exact, identical spawn
transforms every kickoff. Combined with deterministic bot inference
(action_noise = 0 by default), two ships running the same policy from a
mirror-symmetric state produced mirrored, non-diverging play instead of a
real contest — most visible when both sides use the same exported model.
Adds a small position/yaw jitter (well under anything a player would
notice as "not a real kickoff") so kickoff-style resets stop being
bit-for-bit identical.
AIShipController (eval + real gameplay) ran the raw policy output unmasked
regardless of allow_vertical/allow_pitch_roll, while ShipAIController
(training) correctly discarded those axes for grounded curriculum stages.
A grounded-trained model's untrained vertical/pitch-roll output reached the
ship as noise during eval, understating it against models that were never
handicapped this way.
Josh Creek