README.md

ppo_gazebo_tf

Reinforcement Learning for solving the custom cartpole balance problem in gazebo environment using Proximal Policy Optimization(PPO). [Note: this repo is under development]

Environment

• Custom cartpole in gazebo (similar to the one from OpenAI gym)
• Observation Space: 4 (continuos)
• Action Space: 2 (discrete)

Dependencies

• Ubuntu 16.04
• ROS Kinetic
• Gazebo 7
• TensorFlow: 1.1.0
• gym: 0.9.3
• Python 3.6

File setup:

• cartpole_gazebo contains the robot model(both .stl files & .urdf file) and also the gazebo launch file.

• cartpole_controller contains the reinforcement learning implementation of Proximal Policy Optimization(PPO) for custom cartpole environment.

Training Phase:

python3 ppo_train.py

Testing trained policy:

python3 ppo_test.py

References:

• Proximal Policy Optimization by OpenAI
• PPO implementation using TensorFlow

TODO:

• Use Tensorboard for plotting the training and testing graphs.

Project collaborator(s):

Arun Kumar (arunkumar12@iisc.ac.in)