Continuous-PPO

Implementation of the proximal policy optimization on Mujoco environments. All hyper-parameters have been chosen based on the paper.

For Atari domain. look at this.

Demos

Ant-v2	Walker2d-v2	InvertedDoublePendulum-v2

Results

Ant-v2	Walker2d-v2	InvertedDoublePendulum-v2

Dependencies

• gym == 0.17.2
• mujoco-py == 2.0.2.13
• numpy == 1.19.1
• opencv_contrib_python == 3.4.0.12
• torch == 1.4.0

Installation

pip3 install -r requirements.txt

Usage

python3 main.py

• You may use Train_FLAG flag to specify whether to train your agent when it is True or test it when the flag is False.
• There are some pre-trained weights in pre-trained models dir, you can test the agent by using them; put them on the root folder of the project and turn Train_FLAG flag to False.

Environments tested

• [x] Ant
• [x] InvertedDoublePendulum
• [x] Walker2d
• [ ] Hopper
• [ ] Humanoid
• [ ] Swimmer
• [ ] HalfCheetah

Reference

Proximal Policy Optimization Algorithms, Schulman et al., 2017

Acknowledgement

• @higgsfield for his ppo code.
• @OpenAI for Baselines.
• @Reinforcement Learning KR for their Policy Gradient Algorithms.