| Definition | A collection of classic reinforcement learning environments with a variety of tasks and complexities. |
| No of environments | 15 |
Gym Classics is a collection of classic, discrete Markov Decision Processes (MDPs) from reinforcement learning, implemented as OpenAI Gym environments. Gym Classics serves as a benchmark to test agent performance or explore new learning methods.
The library extends the Gym API, offering additional support for dynamic programming, including functionality for model querying. It provides environments with discrete state and action spaces, useful for reinforcement learning and dynamic programming exercises like Q-learning and Value Iteration.
Environments included in Gym Classics cover a variety of domains, each with unique challenges and reward structures. Some environments include ClassicGridworld, CliffWalk, Racetrack, and JacksCarRental, which offer varying levels of complexity, from simple grid-based worlds to more complex problems involving car rental management or racing.