-
The purpose of this repository is study and research about reinforcement learning for robotics control.
-
This repository provides the Model-Free reinforcement learning algorithms.
DDPG
TRPO
PPO
PPO2
SAC
TD3
-
These algorithms are demonstrated in Environment Reacher with ML-Agent.
-
The directory architecture have to be under format.
└─spinning_up_kr
├─env(environment of reacher in unity)
├─mlagents
├─buffer.py
├─core.py
├─ddpg.py
├─ou_noise.py
├─ppo.py
├─ppo2.py
├─sac.py
├─td3.py
└─trpo.py
Reference
[1] Proximal Policy Optimization
[2] High-Dimensional Continuous Control Using Generalized Advantage Estimation
[3] Continuous Control With Deep Reinforcement Learning
[5] Reinforcement Learning Korea PG Travel
[6] Medipixel Reinforcement Learning Repository
[7] Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
[8] tensorflow reinforcement learning framework
[9] Trust Region Policy Optimization
[10] Addressing Function Approximation Error in Actor-Critic Methods