Link to the paper

MuJoCo Experiments

Set-up

1. The experiments have been performed on the following packages:

   • openai gym, version = '0.9.3'
   • MuJoCo mjpro131
   • TensorFlow version 1.10.0

2. Some experiments need changes to be made to the openai gym source code. Copy, paste (, and replace if asked) each and every file inside MuJoCo/gym_files_to_be_merged to their equivalent locations of your gym installation.

Setting up the demonstrator controllers

1. Go to MuJoCo_Demonstrator folder and run the script file generate_demonstrators.sh.

Generating the learners

1. Run the script file overnight.sh present in the root folder.

Visualizing the results

1. Open visualization.ipynb on jupyter notebook and run all cells.