Hopes#

What’s in the box?#

HOPES, which stands for HVAC optimisation with Off Policy Evaluation and Selection, is a Python package for evaluating and selecting RL-based control policies. It offers a set of estimators and tools to evaluate the performance of a target policy, compared to a baseline policy (characterized by an offline logged dataset), using off-policy evaluation techniques. It’s particularly suited for the context of HVAC control, where the target policy is an RL-based controller and the baseline policies are rule-based controllers.

Why Hopes?#

Hopes is designed to be a flexible and easy-to-use package for evaluating and selecting RL-based control policies. Imagine you have a dataset of logged actions and observations from a building HVAC system, and you want to evaluate the performance of one or several RL-based controller. Hopes provides a set of tools to help you do that, including:

  • Estimators for evaluating the performance of a target policy compared to a baseline policy.

  • Tools for selecting the best policy among a set of candidate policies.

  • Tools for visualizing the results of the evaluation and selection process.

  • Dataset preprocessing tools to prepare the data for evaluation.

Installation#

Supported Python versions: 3.10+

From PyPI

1pip install hopes

From source (development version)

1git clone https://github.com/airboxlab/hopes.git
2cd hopes
3# using poetry
4poetry install
5# using pip
6pip install -r requirements.txt