Distributed_neurSLS Documentation

Accompanying code for the paper "Optimal distributed control with stability guarantees by training a network of neural closed-loop maps".

DistributedREN Documentation

Overview

This repository contains the code accompanying the paper titled "Optimal distributed control with stability guarantees by training a network of neural closed-loop maps" authored by Danilo Saccani, Leonardo Massai, Luca Furieri, and Giancarlo Ferrari Trecate.

For inquiries about the code, please contact:

• Danilo Saccani: danilo.saccani@epfl.ch
• Leonardo Massai: l.massai@epfl.ch

Repository Contents

1. main.py: Entry point for training the distributed operator using neural networks.
2. utils.py: Contains utility functions and main parameters for the codebase.
3. models.py: Defines models including the system's dynamical model, Recurrent Equilibrium Network (REN) model, and interconnection model of RENs.
4. plots.py: Includes functions for plotting and visualizing training and evaluation results.

Getting Started

Prerequisites

• Dependencies listed in requirements.txt

Installation

1. Cloning the Repository

git clone https://github.com/DecodEPFL/Distributed_neurSLS.git

2. Navigate to the cloned directory:

cd Distributed_neurSLS

3. Install the required dependencies. We recommend using a virtual environment:

python -m venv venv
source venv/bin/activate  # Activate the virtual environment (Linux/macOS)
pip install -r requirements.txt

Usage

1. Adjust parameters in utils.py as needed.
2. Run the main script to start training:

python main.py

Example

Coordination in mountains problem

The following gifs show trajectories of the vehicles before and after the training of a distributed neurSLS controller, where the agents that need to coordinate in order to pass through a narrow passage while trying to keep a rectangular shape, starting from a random initial position marked with ○.

License

This work is licensed under a Creative Commons Attribution 4.0 International License.