self-organized-control icon indicating copy to clipboard operation
self-organized-control copied to clipboard

To explore how complex and stable motor control can emerge from neurons, we designed neural cellular automata to robustly control a cart-pole agent.

Towards self-organized control

We used neural cellular automata to robustly control a cart-pole agent.

This repository host the interactive article Towards self-organized control as well as the code and a Google Colab notebook to easily reproduce the results and experiment with the pretrained models.

The paper was published in the Innovations in Machine Intelligence (IMI) journal.

Structure

In the code folder:

  • The SelfOrgControl package that host the class and the function to build and run the neural CA. You can install it with pip install git+https://github.com/aVariengien/self-organized-control.git#subdirectory=code
  • The AdditionalExperiments contains code and videos about other experiment with neural CA. Each experiment has its own README.md file.
  • The notebook Towards-self-organized-control-notebook.ipynb
  • The demo contains the javascript code used for the interactive demo using tensorflow.js

Cite this article:

A. Variengien, S. Pontes-Filho, T. E. Glover, S. Nichele, "Towards Self-organized Control: Using Neural Cellular Automata to Robustly Control a Cart-pole Agent", Innovations in Machine Intelligence (IMI), vol. 1, pp. 1-14, 2021. DOI: 10.54854/imi2021.01