Discretized state (bins per dim: (6, 6, 5, 6, 6, 5, 2, 2)), Q-table shape (129600, 4). Eval: mean reward 90.6 +/- 131.8 over 20 episodes.
-
