Inverse Reinforcement Learning Research

This is a research repository, so the code is quite unorganized. Code was adapted from https://github.com/yrlu/irl-imitation and https://github.com/MatthewJA/Inverse-Reinforcement-Learning and converted to PyTorch with a few additions and functions to begin experiemnting with Open-AI gym.

Requirements

• python 3.7
• PyTorch 1.0.0
• Numpy
• cvxopt
• Matplotlib

Files

• Network design is in irl.py
• Main function is in deep_maxent_irl.py

To run gridworld demo with PyTorch

python deep_maxent_irl.py Use -h flag to view gridworld and training parameters

NOTE: CURRENTLY BROKEN WITH PYTHON 3