'pong_video.mp4' video demonstrates the behavior of CC-POMCP with various cost constraint thresholds in Atari 2600 Pong domain.
- When c=200 (unconstrained case): CC-POMCP agent uses the whole map.
- When c=100 (weakly constrained case): CC-POMCP starts to make a trade off between reward and cost.
- When c=30 (strongly constrained case): CC-POMCP is strongly forced to stay in the middle.
