/DDPG

Reimplementing DDPG based on OpenAI Gym + Tensorflow

Primary LanguagePythonMIT LicenseMIT

DDPG

Reimplementing DDPG from Continuous Control with Deep Reinforcement Learning based on OpenAI Gym and Tensorflow

http://arxiv.org/abs/1509.02971

There are still some issues to be solved in this implementation. The performance is still bad. I hope anybody who study DDPG could help figure out how to improve the performance.

How to use

git clone https://github.com/songrotek/DDPG.git
cd DDPG
python gym_ddpg.py

Reference

1 https://github.com/rllab/rllab

2 https://github.com/MOCR/DDPG