donkeycar-rl-autoresearch/agent/experiments/exp11b_parallel_v6.py

"""
Exp 11b: Parallel DummyVecEnv — generated_track + mountain_track, v6 reward.

Changes from Exp 11 (aborted):
  - Reward v6: speed × CTE_quality + efficiency GATE (prevents circular driving)
  - stuck_steps: 80 → 40 (faster termination when stuck against barriers)
  - Everything else identical: same tracks, same hyperparameters, same parallel setup

Setup:
  - Sim 1: 10.0.0.55:9091 → generated_track
  - Sim 2: 10.0.0.55:9093 → mountain_track
  - DummyVecEnv wraps both → PPO sees both tracks in every rollout batch
  - NO env closing, NO set_env(), NO track switching

Hypothesis: v6 reward fixes circular driving exploit seen in Exp 11 while
preserving gradient signal on mountain_track hills. Parallel envs provide
stable multi-track learning (no catastrophic forgetting).
"""
import sys, os, time
sys.path.insert(0, '/home/paulh/projects/donkeycar-rl-autoresearch/agent')

from multitrack_runner import log, StuckTerminationWrapper
from donkeycar_sb3_runner import ThrottleClampWrapper
from reward_wrapper import SpeedRewardWrapper
from stable_baselines3 import PPO
from stable_baselines3.common.vec_env import DummyVecEnv, VecTransposeImage
import gymnasium as gym
import numpy as np

HOST         = '10.0.0.55'
THROTTLE_MIN = 0.2
LR           = 0.000725
TOTAL_STEPS  = 90000
SAVE_DIR     = '/home/paulh/projects/donkeycar-rl-autoresearch/agent/models/exp11b-parallel-v6'
os.makedirs(SAVE_DIR, exist_ok=True)

def make_env(track_id, port):
    def _init():
        raw = gym.make(track_id, conf={'host': HOST, 'port': port})
        env = ThrottleClampWrapper(raw, throttle_min=THROTTLE_MIN)
        env = StuckTerminationWrapper(env, stuck_steps=40, min_displacement=0.5)
        env = SpeedRewardWrapper(env)  # v6: speed×CTE + efficiency gate
        return env
    return _init

log('='*60)
log('Exp 11b: Parallel DummyVecEnv — v6 reward (anti-circle gate)')
log(f'  Sim 1: {HOST}:9091 → generated_track')
log(f'  Sim 2: {HOST}:9093 → mountain_track')
log(f'  throttle_min={THROTTLE_MIN}, lr={LR}, total={TOTAL_STEPS:,}')
log(f'  Reward: v6 (speed × CTE_quality, gated by efficiency >= 0.15)')
log(f'  Stuck: 40 steps (~2.5s)')
log(f'  Method: DummyVecEnv (both tracks in every PPO batch)')
log('='*60)

# Create parallel env
log('Creating DummyVecEnv with two tracks...')
env = DummyVecEnv([
    make_env('donkey-generated-track-v0', 9091),
    make_env('donkey-mountain-track-v0', 9093),
])
env = VecTransposeImage(env)
log(f'  VecEnv num_envs={env.num_envs}, obs={env.observation_space.shape}')

# Create PPO
model = PPO('CnnPolicy', env, learning_rate=LR, verbose=1, device='cpu')
log('PPO created. Starting training...')

# Train in segments for checkpointing
CHECKPOINT_EVERY = 6000
best_reward = float('-inf')
steps_done = 0

while steps_done < TOTAL_STEPS:
    seg_steps = min(CHECKPOINT_EVERY, TOTAL_STEPS - steps_done)
    model.learn(total_timesteps=seg_steps, reset_num_timesteps=False)
    steps_done += seg_steps

    # Save checkpoint
    ckpt = os.path.join(SAVE_DIR, f'checkpoint_{steps_done:07d}')
    model.save(ckpt)
    model.save(os.path.join(SAVE_DIR, 'model'))  # latest for crash recovery
    log(f'[{steps_done:,}/{TOTAL_STEPS:,}] Checkpoint saved: {ckpt}.zip')

    # Quick eval on both tracks simultaneously
    try:
        obs = env.reset()
        ep_rewards = np.zeros(env.num_envs)
        ep_steps = np.zeros(env.num_envs)
        done_mask = np.zeros(env.num_envs, dtype=bool)
        for _ in range(2000):
            action, _ = model.predict(obs, deterministic=True)
            obs, rewards, dones, infos = env.step(action)
            for i in range(env.num_envs):
                if not done_mask[i]:
                    ep_rewards[i] += rewards[i]
                    ep_steps[i] += 1
                    if dones[i]:
                        done_mask[i] = True
            if done_mask.all():
                break

        status0 = '✅' if ep_steps[0] >= 2000 else f'❌@{int(ep_steps[0])}'
        status1 = '✅' if ep_steps[1] >= 2000 else f'❌@{int(ep_steps[1])}'
        log(f'  Eval: gen_track={ep_rewards[0]:.1f}r/{int(ep_steps[0])}s {status0}  '
            f'mountain={ep_rewards[1]:.1f}r/{int(ep_steps[1])}s {status1}')

        total_reward = ep_rewards.sum()
        if total_reward > best_reward:
            best_reward = total_reward
            model.save(os.path.join(SAVE_DIR, 'best_model'))
            log(f'  ⭐ NEW BEST: {best_reward:.1f} (combined)')
    except Exception as e:
        log(f'  Eval error: {e}')
        import traceback; traceback.print_exc()

model.save(os.path.join(SAVE_DIR, 'model'))
log(f'\nTraining complete. Best combined reward: {best_reward:.1f}')
log(f'Checkpoints in {SAVE_DIR}:')
for f in sorted(os.listdir(SAVE_DIR)):
    size = os.path.getsize(os.path.join(SAVE_DIR, f)) // (1024*1024)
    log(f'  {f}  ({size}MB)')

# Close training env
env.close()
time.sleep(5)

# --- Eval on all 4 tracks using sim 1 (port 9091) ---
# We use a single sim for sequential eval since we only need one track at a time
log('\n' + '='*60)
log('EVALUATION: best_model on 4 tracks (3 sets each)')
log('='*60)

EVAL_TRACKS = [
    ('donkey-mountain-track-v0',   'mountain_track'),
    ('donkey-generated-track-v0',  'generated_track'),
    ('donkey-generated-roads-v0',  'generated_road'),
    ('donkey-minimonaco-track-v0', 'mini_monaco'),
]
EVAL_SETS = 3
EVAL_MAX_STEPS = 2000
EVAL_PORT = 9091

best_model_path = os.path.join(SAVE_DIR, 'best_model.zip')
results_by_track = {}

for track_id, track_name in EVAL_TRACKS:
    log(f'\n--- {track_name} ---')
    steps_list = []
    
    for s in range(1, EVAL_SETS + 1):
        try:
            # Create fresh env for each eval run
            raw = gym.make(track_id, conf={'host': HOST, 'port': EVAL_PORT})
            eval_env_inner = ThrottleClampWrapper(raw, throttle_min=THROTTLE_MIN)
            eval_env_inner = StuckTerminationWrapper(eval_env_inner, stuck_steps=40, min_displacement=0.5)
            eval_env_inner = SpeedRewardWrapper(eval_env_inner)
            eval_env = VecTransposeImage(DummyVecEnv([lambda e=eval_env_inner: e]))
            
            eval_model = PPO.load(best_model_path, env=eval_env, device='cpu')
            
            obs = eval_env.reset()
            total_r, total_s, done = 0.0, 0, False
            while not done and total_s < EVAL_MAX_STEPS:
                action, _ = eval_model.predict(obs, deterministic=True)
                result = eval_env.step(action)
                if len(result) == 4:
                    obs, r, d, info = result
                    done = bool(d[0])
                else:
                    obs, r, t, tr, info = result
                    done = bool(t[0] or tr[0])
                total_r += float(r[0])
                total_s += 1
            
            status = '✅' if total_s >= EVAL_MAX_STEPS else f'❌@{total_s}'
            log(f'  Set{s}: {total_r:.1f}r / {total_s}s {status}')
            steps_list.append(total_s)
            
            eval_env.close()
            time.sleep(3)
        except Exception as e:
            log(f'  Set{s}: ERROR — {e}')
            steps_list.append(0)
            time.sleep(3)
    
    mean_steps = np.mean(steps_list) if steps_list else 0
    results_by_track[track_name] = steps_list
    log(f'  Mean: {mean_steps:.0f} steps')

# Summary
log('\n' + '='*60)
log('SUMMARY')
log('='*60)
for track_name, steps_list in results_by_track.items():
    steps_str = '/'.join(str(s) for s in steps_list)
    mean = np.mean(steps_list)
    verdict = '✅' if mean >= 1500 else '⚠️' if mean >= 500 else '❌'
    log(f'  {verdict} {track_name:20s}: {steps_str}  mean={mean:.0f}')

log(f'\n=== Exp 11b COMPLETE ===')