Reinforcement Learning Environments

Below you find a list of available reinforcement learning environments, mostly build on top of Gym.
(See a long curated list here as well)

Intended for: BSc, MSc, PhD

Core:

• Gymnasium

  • Atari

  • Mujoco

  • Toy Text

  • Classic Control

  • Box2D

Robotics:

• Gym-Robotics

• PyBullet Gym

• RoboGym

• SpaceRobot

Simple games:

• Flappy Bird

• Slime Volley

• Highway

• PyGame

• Maze

Multi-agent:

• PettingZoo (multi-agent)

Offline RL (datasets):

• D4RL

• Minari

Meta-RL:

• MetaWorld (meta RL)

Micro (complex environments with relatively low dimensionality):

• MiniGrid

• MiniWorld

• MicroRTS

• MinAtar

• Crafter (Cowherd, BridgeWalk)

Procedurally generated:

• Procgen (rewrite)

• Neural MMO

• Unity environments

• Retro

• DM Control Suite

• Sokoban

Racing

• RaceCar

• TORCS racecar

• Learn2Race

Robustness/Safety:

• Safety Gym

• Anomalous RL environments

Object-centric learning:

• Spriteworld

• Block Moving

Benchmarking:

• CARL

• Behavioural Suite

Self-driving car:

• CARLA

• DuckyTown
  

Energy management:

• Bauwerk

Board games:

• Go

• Chess

• Hex

• Rubiks Cube

First-person navigation:

• Doom

• Minecraft

• Habitat

• Obstacle tower

• MarsExplorer

Multi-objective:

• MO-Gym

Traffic simulation:

• Flow (traffic simulation)

• Sumo RL (traffic simulation)