FrozenLake Reinforcement Learning Environment with Slippery Dynamics

Derived from the FrozenLake-v1 reinforcement learning environment with the 'slippery' parameter set to True. It contains text-based state and action data for training and evaluating RL agents. The author is AntoineLB, and it was last updated in April 2022.

Use Cases

• Train a Q-learning agent to navigate the grid using state and action sequences.
• Benchmark policy gradient methods on the stochastic transitions inherent in the slippery dynamics.
• Analyze agent performance and failure modes within the text-based environment representation.

Strengths

• Based on a standard, well-defined RL benchmark environment (FrozenLake-v1).
• Explicitly configures the environment's stochastic 'slippery' dynamics for consistent evaluation.
• Hosted on a major platform (Hugging Face) with a recorded update date (2022-04-22).

Limitations

• The specific data format, size, and column structure are unknown from the input.
• The dataset's recency may be a limitation for applications requiring the latest environment versions.
• Lack of detailed documentation on the exact content and sampling methodology.

Provenance

Source

Hugging Face dataset repository, author AntoineLB.

Collection Method

Derived from the OpenAI Gym FrozenLake-v1 simulation environment.

Freshness

Last updated on 2022-04-22.