01dbfc7ede
feat(*): Add exported Linux binary training path for faster parallel instances
2026-07-28 20:56:14 +01:00
ac317cd2e1
feat(*): Add sudden-death overtime with golden goal on a draw
2026-07-28 19:44:35 +01:00
09ea8d6fbc
feat(*): Style the results screen and dock the heading tape under the scoreboard
2026-07-28 19:32:40 +01:00
a9b7b450d5
feat(*): Redesign HUD scoreboard as a single timer/score/team banner
2026-07-28 19:24:40 +01:00
93f90ca6de
feat(*): Add 3-2-1 kickoff countdown before play and after goals
2026-07-28 19:11:36 +01:00
a0b527b8cb
chore(training): Add 20260726-1904-curric-s1-unmask checkpoints, logs, and exported policy
2026-07-28 00:29:06 +01:00
31aa5cae8c
chore(*): Remove superseded pre-curriculum and dead-branch runs
2026-07-26 19:25:22 +01:00
f7d03b1e74
chore(*): Remove superseded curriculum branches
2026-07-26 19:19:01 +01:00
653afe4040
chore(training): Add curric-s6-unmask-retry2 checkpoints, logs, and exported policy
2026-07-26 13:06:31 +01:00
4e0efec612
chore(training): Add curric-s6-unmask-retry1 checkpoints, logs, and exported policy
2026-07-25 07:39:52 +01:00
580222c139
feat(*): Promote curric-s6-unmask as the shipped "easy" bot
2026-07-24 09:18:53 +01:00
f4dfc43d04
chore(training): Add curric-s6-unmask checkpoints, logs, and exported policy
2026-07-24 02:23:34 +01:00
1afdc301ab
feat(training): add airborne_penalty and a stage-6 "unmask" curriculum run
...
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.
2026-07-22 21:22:27 +01:00
fda784d50d
chore(training): Add curric-s5-aggression checkpoints, logs, and exported policy
2026-07-22 20:31:54 +01:00
0e42182cce
fix(*): jitter kickoff resets to break deterministic same-model mirror matches
...
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.
2026-07-22 18:01:21 +01:00
cf4859e61c
chore(training): Add curric-s4-mechanics checkpoints, logs, and exported policy
2026-07-22 11:53:29 +01:00
8c15c466ef
fix(*): apply the locomotion mask during in-game/eval inference, not just training
...
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.
2026-07-21 22:23:09 +01:00
bc644e0305
chore(training): Add curric-s3-no_draws checkpoints, logs, and exported policy
2026-07-21 21:44:51 +01:00
ffa9017cbd
chore(training): Add curric-s2-defend checkpoints, logs, and exported policy
2026-07-21 19:17:32 +01:00
523a26aa60
chore(training): Add curric-s1-score checkpoints, logs, and exported policy
2026-07-21 16:50:08 +01:00
8e3fafcc8b
feat(*): add staged curriculum training with an automated stage-by-stage orchestrator
2026-07-21 12:38:51 +01:00
1d539cc8c7
chore(training): Add run11 checkpoints, logs, and exported policy
2026-07-21 09:40:13 +01:00
fbf51eebe5
chore(training): Add run10 checkpoints, logs, and exported policy
2026-07-20 22:34:32 +01:00
7d69ac4a01
feat(*): retune scoring incentives and add finishing reps
2026-07-20 20:06:11 +01:00
46c8275523
feat(*): add wall/ceiling surface pull and retune ball-ship materials
2026-07-20 19:33:28 +01:00
e23d780402
chore(training): Add run09 checkpoints, logs, and exported policy
2026-07-20 19:09:32 +01:00
3457d4ca84
feat(*): Add rounded arena boundaries and reward shaping to curb corner-camping
2026-07-20 08:20:33 +01:00
240e362f2f
feat(*): Add bot selection, score HUD, and winner reveal to matches, replace HUD text with aircraft-style flight instruments, and fix camera judder
2026-07-20 07:12:09 +01:00
48369c50dc
chore(training): Add run07 checkpoints, logs, and exported policy
2026-07-20 05:56:00 +01:00
b8f5f4938d
chore(training): Add run06 checkpoints, logs, and exported policy
2026-07-20 00:42:52 +01:00
4f075b352f
feat(*): Add orientation-readable team-coloured ship meshes, smooth the ball-cam orbit, and hide arena walls the camera is outside of
2026-07-19 22:18:21 +01:00
dfb7285704
chore(training): Add run05 checkpoints, logs, and exported policy
2026-07-19 21:58:08 +01:00
cd94252f92
chore(*): Remove run05 artifacts trained with collapsed exploration
2026-07-19 19:16:01 +01:00
c517a85508
chore(training): Add run05 checkpoints, logs, and exported policy
2026-07-19 19:11:31 +01:00
4c227c24b4
feat(*): Add a distance-to-ball penalty so idling scores negative, boost ball-seeking rewards, and halve the wall-contact penalty
2026-07-19 18:04:36 +01:00
80e9b676d5
chore(training): Add run04 checkpoints, logs, and exported policy
2026-07-19 17:54:36 +01:00
c8f052854c
fix(*): Detect wall contact by contact normal so floor contact anywhere on the pitch is exempt from the wall penalty
2026-07-19 15:53:30 +01:00
4e2d406aa2
chore(*): Remove run03 artifacts trained against the floor-taxed reward
2026-07-19 15:35:55 +01:00
7777280062
feat(*): Exempt the floor from the wall-contact penalty, add a tilt penalty for non-upright flight, and double velocity-to-ball shaping
2026-07-19 15:34:34 +01:00
5973d0893b
chore(training): Add run03 checkpoints, logs, and exported policy
2026-07-19 15:21:06 +01:00
772f98b7fe
feat(*): Fix exported-policy action order to gymnasium's sorted-key layout, add wall-contact penalty and stronger ball-touch reward, and wire Spectate to run01 vs run02
2026-07-19 13:21:14 +01:00
6f1a840249
chore(training): Add run02 checkpoints, logs, and exported policy
2026-07-19 12:41:44 +01:00
07217c3517
feat(*): Add bot-vs-bot Spectate mode with main-menu entry, entropy-control flags (--ent-coef, --reset-std) for resumed training runs, and a Linux/3090 remote-training guide (TRAINING_LINUX.md)
2026-07-19 10:10:32 +01:00
379ef9910e
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
2026-07-18 20:18:03 +01:00
85f96eb15e
feat(*): Add self-play RL training pipeline with PPO trainer, in-game GDScript policy inference, and bot opponent support in Match mode
2026-07-18 19:32:51 +01:00
328831df1f
refactor(*): Restructure game into reusable Arena/GameMode architecture with controller-driven ships, adding Free Play and Match modes
2026-07-18 15:34:11 +01:00
14a8907038
build(*): Update project to Godot 4.7
2026-07-18 14:49:23 +01:00
c3329b3280
feat(*): Add goal
2025-07-20 20:51:24 +01:00
8650995c18
chore(*): Remove redundant files
2025-07-16 23:05:32 +01:00
35a23f79e3
refactor(*): Make hud more performant and maintainable
2025-07-16 19:19:58 +01:00
Josh Creek