DDPG + HER

Requirment

• python
• keras
• Mujoco-py
• NumPy
• gym

Test Envirment

• Ubuntu 2004 LST
• python 3.8
• neural network update time 200 * 50 * 50
• initial state is limited with distance close to 0.05 (just for efficiency)
• openai environment FetchPickAndPlace-v1

How to use

How to train

• change the configuration variable PLAY_MODEL come from main.py to False
• Run main.py

How to play

• Change the configuration variable PLAY_MODEL come from main.py to True
• Run main.py

Model Explain

Combination with Deep Deterministic Policy Gradient(DDPG) and Hindsight Experience Replay(HER) algorithm.