reward ajustments

robot starts balancing without changing phase and farms alive bonus
fix ->external force and bigger height penalty

ml/train.py

@@ -189,8 +189,8 @@ def train(
         verbose=1,
         seed=seed,
         learning_rate=1.5e-4,      # Cut LR in half (from 3e-4) to smooth out updates
-        n_steps=1024,              # Larger rollout buffer per env for stable gradients
-        batch_size=128,            # Larger minibatches reduce noise
+        n_steps=256,               # Larger rollout buffer per env for stable gradients
+        batch_size=256,            # Larger minibatches reduce noise
         n_epochs=10,               # Number of epoch updates per rollout
         gamma=0.99,                # Discount factor
         gae_lambda=0.95,           # GAE smoothing