RL_MUJOCO

Example

Back propagation using PPO

Kolen Pollack using PPO

Hebbian using PPO

Purpose

RL-MUJOCO uses optimization-based algorithms such as SAC, DDPG, and PPO, as well as bio-plausible algorithms such as Hebbian PPO and Kolen-Pollack PPO to simulate multiple environments within the gym.

Files and Directories

config.ini

Configuration file for setting parameters and hyperparameters for the training process.

main.py

The main script to start training and evaluating the agents.

agents

Contains implementations of various reinforcement learning algorithms.

• __init__.py: Initialization file for the agents module.
• ddpg.py: Implementation of the Deep Deterministic Policy Gradient (DDPG) algorithm.
• Hebbianppo.py: Implementation of the Hebbian Proximal Policy Optimization (PPO) algorithm.
• ppo.py: Implementation of the Proximal Policy Optimization (PPO) algorithm.
• sac.py: Implementation of the Soft Actor-Critic (SAC) algorithm.

log

Contains logs of training runs for different environments and algorithms.

model_weights

Contains saved model weights for different environments and algorithms.

runs

Contains TensorBoard log files for visualizing training metrics.

utils

Contains utility scripts for noise generation and other helper functions.

• __init__.py: Initialization file for the utils module.
• noise.py: Functions for noise generation used in exploration.
• utils.py: Miscellaneous utility functions.

Usage

1. Setup: Ensure you have the required dependencies installed.

   pip install -r requirements.txt
   

Implement

You can write the code in the terminal and execute it

python main.py --env_name Ant-v4 --train True --tensorboard True --algo ppo  

Reference

• https://github.com/seolhokim/Mujoco-Pytorch.git