Read my blog to see this project in action!!!
Reinforcement Learning(RL) has sparked the interest of awesome minds around the world and as a result huge advacements have been made in the last decade, but mostly in applications in which we can afford to fail quickly and safely such as videogames or robotics (to a lesser extend).
Nowadays, researchers have turned their attention to domains where safety is a top priority. Examples of those applications are almost everywhere, from self-driving cars and unmanned aerial vehicles to use cases in the classical manufacturing industries: Chemical, manufacturing.
This repository is my contribution towards spreading RL in the Chemical Engineering field. Concretely, I have layed here the foundations of what I envision to be a library that would help researchers in this area to run quick experiments, compare results and even, develop new techinques. In short, like OpenAI Gym, but for Chemical Engineering.
To my knowledge, no library of this kind has been proposed yet, although it is very likely that other people is already working in something similar.
At the moment, only commonly used algorithms and techniques are available, (Deep Q Network, experience replay, tabular methods), with time, more environments and algorithms will be added. I am currently working on one implementation of safe RL based on Gaussian Processes, this addition will be the first step towards the introduction of safe RL techniques to the library which, in the worst case, would serve as examples for the newcomers. The latter proposition takes care of the need of reliable and clear resources about Safe RL, something that I would have liked to have when I started to learn about it.
The long term vision of this library is to serve as a reference of RL in chemical engineering and a vehicle to distribute the breakthroughs of RL in Chemical Engineering among the ML community.
If you find an error or would like to contibute, please reach out!.