https://github.com/befelix/safe-exploration

Safe Exploration with MPC and Gaussian process models
https://github.com/befelix/safe-exploration

exploration model-predictive-control reinforcement-learning safety

Last synced: 7 days ago
JSON representation

Safe Exploration with MPC and Gaussian process models

• Host: GitHub
• URL: https://github.com/befelix/safe-exploration
• Owner: befelix
• License: mit
• Created: 2018-12-10T10:34:11.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2020-08-17T17:33:47.000Z (about 5 years ago)
• Last Synced: 2025-08-21T01:39:12.551Z (about 2 months ago)
• Topics: exploration, model-predictive-control, reinforcement-learning, safety
• Language: Python
• Homepage:
• Size: 573 KB
• Stars: 89
• Watchers: 2
• Forks: 23
• Open Issues: 0
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE

Awesome Lists containing this project

README

          
=============================================================

 Learning-based Model Predictive Control for Safe Exploration

=============================================================

.. image:: https://circleci.com/gh/befelix/safe-exploration/tree/master.svg?style=svg

    :target: https://circleci.com/gh/befelix/safe-exploration/tree/master

    

.. image:: https://travis-ci.com/befelix/safe-exploration.svg?branch=master

    :target: https://travis-ci.com/befelix/safe-exploration

This code accompanies the following paper:

.. [1] T. Koller, F. Berkenkamp, M. Turchetta, A. Krause,

  `Learning-based Model Predictive Control for Safe Exploration `_

  in Proc. of the Conference on Decision and Control (CDC), 2018

Installation

------------

Install the library including all dependencies with.

::

  pip install -e ".[test,visualization,ssm_gpy,ssm_pytroch]"

  

  

`test` for the testing tools. `visualization` for visualizations such as matplotlib. `ssm_gpy` and `ssm_pytorch` for state space models based on `GPy` or `PyTorch`, respectively.

Experiments can be run using the `experiments/run.py` script.

Test can be run using pytest. There are also more sophisticated style tests in 

`scripts/test_code.sh`.