/gym-adv

Gym environments modified with adversarial agents

Primary LanguagePython

Under Development

Gym environments with adversarial disturbance agents

This contains the adversarial environments used in our work on Robust Adversarial Reinforcement Learning (RARL). We heavily build on OpenAI Gym.

Getting Started

The environments are based on the MuJoCo environments wrapped by OpenAI Gym's environments (info). For more information on OpenAI Gym environments refer to the Gym webpage.

Since these environments use the OpenAI pyhton bindings for the MuJoCo environments, you'll need to install mujoco-py following this.

Example

import gym
E = gym.make('InvertedPendulumAdv-v1')
current_observation = E.reset()

# Set maximum adversary force
E.update_adversary(6)

# Get a sample action
u = E.sample_action()
# u.pro corresponds to protagonist action, while u.adv corresponds to the adversary's action

# Perform action 
new_observation, reward, done, ~ = E.step(u)

Contact

Lerrel Pinto -- lerrelpATcsDOTcmuDOTedu.