Agent Training & Evaluation Environments

21 resources

• Acrobot-v1

  A two-link swing-up benchmark that tests control when only one joint is actuated.

  Farama Foundation · Local Gymnasium environment. No account required.
• Bipedal Walker

  A continuous-control benchmark for moving a two-legged simulated robot across terrain.

  Farama Foundation · Local Gymnasium environment. No account required.
• Blackjack-v1

  A simulated blackjack benchmark for testing decisions under uncertainty with a small observation space.

  Farama Foundation · Local Gymnasium simulation. No account required.
• Car Racing

  An image-based driving task that tests steering and speed control on generated circuits.

  Farama Foundation · Local Gymnasium environment. No account required.
• CartPole-v1

  A small control benchmark where an agent balances a pole by pushing its cart left or right.

  Farama Foundation · Local Gymnasium environment. No account required.
• CliffWalking-v1

  A grid navigation benchmark for comparing short routes with routes that avoid a costly cliff edge.

  Farama Foundation · Local Gymnasium environment. No account required.
• FrozenLake-v1

  A compact grid navigation benchmark for reaching a goal while avoiding holes and handling slippery movement.

  Farama Foundation · Local Gymnasium environment. No account required.
• Gymnasium

  A local Python interface for training and testing reinforcement learning agents in simulated environments.

  Farama Foundation · Local Python package. No account required.
• Lunar Lander

  A physics benchmark for bringing a lander to a target with discrete thrust commands.

  Farama Foundation · Local Gymnasium environment. No account required.
• MiniGrid DoorKey 5×5

  A grid-world benchmark for learning a sequence of actions involving a key, a locked door, and a goal.

  Farama Foundation · Local MiniGrid environment using Gymnasium. No account required.
• MountainCar-v0

  A momentum benchmark where an underpowered car must build speed before climbing a hill.

  Farama Foundation · Local Gymnasium environment. No account required.
• Pendulum-v1

  A continuous-control benchmark for swinging a pendulum upright and keeping it there.

  Farama Foundation · Local Gymnasium environment. No account required.
• PettingZoo Chess

  A two-player chess environment for testing legal-action handling and strategic policies.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Connect Four

  A two-player game benchmark for evaluating move selection, opponent modeling, and search.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Cooperative Pong

  A visual coordination task in which two agents work together to keep a ball in play.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Pistonball

  A cooperative physics benchmark where agents control pistons to move a ball across the screen.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Pursuit

  A cooperative grid benchmark where multiple pursuers coordinate to capture moving evaders.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Rock Paper Scissors

  A repeated two-player game for evaluating mixed strategies and adaptation to an opponent.

  Farama Foundation · Local PettingZoo environment. No account required.
• PettingZoo Tic-Tac-Toe

  A two-player benchmark for testing turn handling, legal moves, and simple game policies.

  Farama Foundation · Local PettingZoo environment. No account required.
• Reacher

  A MuJoCo benchmark in which a two-joint arm moves its fingertip toward a target.

  Farama Foundation · Local Gymnasium environment. No account required.
• Taxi

  A discrete routing task that combines navigation with passenger pickup and drop-off.

  Farama Foundation · Local Gymnasium environment. No account required.