Solving OpenAI Gym Environments

A PyTorch Implementation

OpenAI Gym Environment

Content

Environment	DQN	Double-DQN	Dueling-DQN	REINFORCE	SPS
CartPole-v1	⏳	⏳	⏳	☑️	☑️

• 💦 work in progress
• ⏳ hyperparameter tuning
• ☑️ done

Todo

Classic Control

• Acrobot
• CartPole
• MountainCar
• MountainCar Continuous
• Pendulum

MuJoCo: Continuous Control

• Ant
• HalfCheetah
• Hopper
• Humanoid
• Humanoid Standup
• Inverted Pendulum
• Inverted DoublePendulum
• Reacher
• Swimmer
• Walker 2D

Box2D: Continuous Control

• Bipedal Walker
• Bipedal Walker Hardcore
• Car Racing
• Lunar Lander
• Lunar Lander Continuous

Robotics: Simulated Goal-Based Tasks

• Fetch Pick and Place
• Fetch Push
• Fetch Reach
• Fetch Slide
• Hand Manipulate (Block)
• Hand Manipulate (Egg)
• Hand Manipulate (Pen)
• Hand Reach

Environment

Usage

Add the root directory to PYTHONPATH

export PYTHONPATH=$PWD

Reference

Citation