Views
No views yet
| Metric | Value |
|---|---|
| Algorithm | SAC (Soft Actor-Critic) |
| Training steps | 1.91M |
| Training time | ~60 min (MacBook M-series, CPU) |
| Parallel envs | 8 |
| Network | MLP [256, 256] |
| Best reward | 530 |
| Mean distance | 2.65m |
| Episode length | ~200/1,000 (~4 seconds upright) |
| Status | Balancing + stumbling forward |
1from stable_baselines3 import SAC
2
3model = SAC.load("best/best_model")
4
5obs, _ = env.reset()
6for _ in range(1000):
7 action, _ = model.predict(obs, deterministic=True)
8 obs, reward, done, truncated, info = env.step(action)reward = forward_vel × 5.0 # primary: move forward
+ alive_bonus × 1.0 # stay upright
+ upright_reward × 0.3 # orientation bonus
- ctrl_cost × 0.001 # minimize energy
- lateral_penalty × 0.3 # don't drift sideways
- smoothness × 0.0001 # discourage jerky motionbest/best_model.zip — Best checkpointcheckpoints/ — All 100K-step checkpointslogs/evaluations.npz — Evaluation metricsg1_balancing.mp4 — Demo video