ndroi / inverse_rl

Inverse RL

Implementations for imitation learning / IRL algorithms in RLLAB

Contains:

• GAIL (https://arxiv.org/abs/1606.03476/pdf)
• Guided Cost Learning, GAN formulation (https://arxiv.org/pdf/1611.03852.pdf)
• Tabular MaxCausalEnt IRL (http://www.cs.cmu.edu/~bziebart/publications/thesis-bziebart.pdf)

Setup

This library requires:

• rllab (https://github.com/openai/rllab)
• Tensorflow

Examples

Running the Pendulum-v0 gym environment:

1. Collect expert data

python scripts/pendulum_data_collect.py

You should get an "AverageReturn" of around -100 to -150

2. Run imitation learning

python scripts/pendulum_gcl.py

The "OriginalTaskAverageReturn" should reach around -100 to -150