EX2 Exloration with Exemplar Models for Deep Reinforcement Learning

This repository contains the code to run the experiments from this paper.

Installation

Dependecies

• rllab
• Mujoco v1.31
• OpenAI Gym using commit hash 518f4b
• gym-doom

Once rllab is installed run the following command in the root folder of rllab:

git submodule add -f git@github.com:jcoreyes/ex2.git

For doom, you need to copy ex2/ens/dooms/assets/my_way_home2.wad to <doom_py-package-location>/scenarios/my_way_home2.wad. .

EC2 Setup

Except for doom, most experiments are meant to be run on EC2 but can also be run locally. Follow these instructions to setup EC2. Furthermore the information in misc/ec2_info will need to configured according to your AWS security settings.

Running experiments

Then you can run ex2 on the appropiate environment via python ex2/exps/pointmass.py. Change the mode variable in the script to 'ec2' to run on EC2.