A collection of Reinforcement Learning code resources, frameworks and environments. Every link below is a GitHub link so this is mainly focused on resources that are directly related to open-source code. For other types of resources check out awesome-rl, spinning up and our 4 hour Reinforcement Learning course we created: An Introduction to Deep Reinforcement Learning
Feel free to raise issues, pull requests or email us at [email protected] for missing resources you think we should add.
There really aren't many truly popular RL "frameworks" (i.e. the equivalent of DL frameworks but for RL) and there isn't even agreed upon formal design patterns/workflows that are common across all RL algorithm implementations. Also, the distinction between framework and "collection of algorithms" is hard to specify sometimes. Therefore, the below list is a mix within this continuum as well as some RL courses.
- pytorch-rl
- PyTorch-RL
- adeptRL
- torch-rl
- torchrl
- rlkit
- SLM-Lab
- salina
- pfrl
- tianshou
- lagom
- minimalRL
- RLHive
- mbrl-lib
- DeepRL-Tutorials
- ikostrikov's a3c, a2c, ppo, acktr, trpo
- higgsfield's RL Adventure 1 and 2: DQN and Policy Gradient variations
- Deep-Reinforcement-Learning-Algorithms-with-PyTorch
- reinforcement-learning-algorithms
- vel
- rl_algorithms
- practical_RL but other frameworks too (TF, theano, lasagne)
- Udacity deep-reinforcement-learning
- simoninithomas / Deep_reinforcement_learning_Course
- Deep-Reinforcement-Learning-Hands-On
- ShangtongZhang/DeepRL
- ptan
- autonomous-learning-library
- AGNES
- rltime
- rlpyt
- cleanrl
- machina
- cherry
- yarr
- pytorch-madrl
- torch_rl
- Deep_RL_with_pytorch
- Catalyst.RL
- pg_travel
- spinning-up-basic PyTorch version of OpenAI's spinning up
- firedup spinning up PyTorch
- humblerl
- ReAgent renamed from Horizon
- OpenAI baselines
- stable-baselines
- coach
- yarll
- acme
- dopamine
- tensorforce
- trfl
- huskarl
- rl_algorithms
- Robotics-rl-srl
- Tidy-Reinforcement-learning
- reinforcement-learning denny britz
- OpenAI Spinning Up
- agents
- tensorflow-reinforce
- DeepRL-Agents
- anyrl-py
- reinforcement_learning
- Reinforcement-learning-with-tensorflow
- basic_reinforcement_learning
- mushroom
- reaver
- rllab
- garage
- reinforcejs
- rlpy
- keras-rl
- mrl
- tonic
- magi
- prl
- kerlym
- godot_rl_agents
- rlax
- rlstructures
- rl-agents
- simple_rl
- ShangtongZhang/reinforcement-learning-an-introduction
- reinforcement-learing implementations
- reinforcement-learning-kr
- PARL
- DeeR
- sinergym
- OpenAI Gym. Contains Algorithmic, Atari, box2d, classic control, MuJoCo, Robotics, Toy text and many more environments.
- OpenAI Universe which was archived and muniverse
- OpenAI Retro
- ml-agents
- PyGame-Learning-Environment
- MinAtar
- toybox
- PCSX-Reloaded
- rlcard
- metaworld
- open_spiel
- MAMEToolkit
- Arena-BuildingToolkit
- CHALET
- House3D
- ai2thor and RoboTHOR
- manipulathor
- MINOS
- ViZDoom
- diy-gym
- dm_hard_eight
- DeepMind Lab and psychlab
- HoME
- Malmo, marLo and gym-minecraft
- AirSim
- CARLA
- UnrealROX, The RobotriX and gym-unrealcv
- holodeck
- marathon-envs
- ToriLLe
- GibsonEnv and GibsonEnvV2 and iGibson
- obstacle-tower-envil
- DeepGTAV
- virtualhome
- deepdrive
- gym_torcs
- neural-mmo
- SelfDrive_AI
- SEVN
- tmrl
- rocket-league-gym
- habitat-sim and habitat-api
- angry-ai
- gym-miniworld
- gym-duckietown
- dm_alchemy
Some of these aren't necessarily for RL but could be for motion planning and simulating complex robotics systems.
- pybullet-gym and official examples at bullet3/examples/pybullet and pybullet_robots
- dm_control
- roboschool
- robosuite
- RLBench
- leibnizgym
- furniture
- roboverse
- SAPIEN-Release
- Unity-Robotics-Hub
- mujoco
- gym-gazebo and ign-gazebo
- gym-gazebo2
- cassie-rl and cassie-mujoco-sim
- osim-rl
- raisimLib
- SimuRLacra
- morse
- graspit
- CausalWorld
- flatland
- gym-fetch
- brax
- isaac-gym not GitHub but I'll make an exception. https://github.com/NVIDIA-Omniverse/IsaacGymEnvs
- panda-gym
- tactile_gym
- tacto
- ostrichrl
- nimblephysics
- assistive-gym
- ravens
- argos3
- snakebot-gym
- dedo
- realworldrl_suite
- simulation-construction-set
- rex-gym
- jiminy
- racing-rl
- PackIt
- Stage
- robogym
- choreonoid
- Aslan
- pyphysx
- robo-gym
- flightmare
- gym-pybullet-drones
- gym-saywer
- spot_mini_mini
- webots
- SenseAct
- offworld-gym
- gym-electric-motor
- webots
- gym-ignition
- rl_deepracer_robomaker_coach_gazebo
- Klampt
- gym-chrono
- multiworld
- autorally
- FlightGoggles
- v_rep and PyRep
- pycolab
- escape_room
- MazeBase
- lab2d
- spriteworld
- gym-maze
- gym-sokoban
- mazelab
- GridUniverse Created by us
- gym-minigrid
- hanabi-learning-environment
- bsuite
- minihack
- XWorld 2D+3D
- phyre
- tdw
- gym-racecar
- phy-q/benchmark
- CompilerGym
- sumo-rl
- AutonomousBlimpDRL
- Rocket_Lander_Gym
- megastep
- nle
- balloon-learning-environment
- magical
- SIMPLE
- alfworld
- ElderSim
- SoftGym
- gym-cooking
- rl_unplugged
- flow
- lf2gym
- safety-gym
- assistive-vr-gym
- highway-env
- gym-super-mario-bros and gym-super-mario
- Trading-Gym and gym-trading
- gymfc