CorteXforge

CorteXforge is an end-to-end framework designed to automate the generation and execution of radio dataset experiments on the SLICES-RI/CorteXlab testbed. It relies on the GNU Radio environment to record labeled transmissions of various signals.

Overview

This project is organized into three main components:

• Scenario generation: this part produces configuration files describing the experiment setup. It creates:
  • a scenario.yaml file defining which nodes will be used on CorteXlab;
  • an timeline.csv file orchestrating the role and sequence of these nodes.
• Experiment execution: this part deploys and executes the generates experiment definitions (timeline.csv) directly on the SLICES-RI/CorteXlab nodes.
• Dataset API

Quick start (User Guide) 🚀

1. Scenario Generator

⚠️ This module is implemented in Python 3.13 !

The scenario generator can be executed locally before deployment in Slices/CorteXlab. It allows configuration of experimental parameters such as:

• selected nodes to be used
• time of recording (in seconds)

Example usage

• git clone https://github.com/Andreaj42/CorteXForge.git
• python3.13 -m venv .venv
• . .venv/bin/activate
• pip install -e .[planner]
• cortexforge-planner --nodes-path confis/nodes.yaml --duration 600 --output-path my/path/on/cortexlab

2. Forge (Experiment Execution)

⚠️ This part must be executed directly on the Slices/CorteXlab testbed !

Each nodes defined before in the previous stage will run a GNU Radio flowgraph according to the configuration.

Example usage

First, connect to the testbed:

• ssh username@gw.cortexlab.fr

Next, book the testbed with your selected nodes (nodes: 5, 10, and 31 here) for the time of recording (increase the value):

• oarsub -l {"network_address in ('mnode5.cortexlab.fr', 'mnode10.cortexlab.fr', 'mnode31.cortexlab.fr')"}/nodes=3,walltime=0:20:00 -r "2025-10-12 21:03:00"

To delete a job, use:

• oardel job_id

And move the previously generated experiment folder into your Cortexlab home, then run:

• minus task create experiment -f
• minus task submit experiment.task

To monitor your experiment, use:

• minus testbed status
• minus log -d

3. Pre-generated Datasets

Just want to use a dataset without running your own experiments? Simply download one of the pre-generated datasets produced with CorteXforge.

First, install the dataset API:

• pip install cortexforge

List all available datasets:

• cortexforge datasets list

Download a dataset:

• cortexforge datasets download <dataset_name>

For example:

• cortexforge datasets download modfore

Docker Images 🐳

To simplify deployment and ensure reproductibility, we generated a Docker image. This image extends the standard CorteXlab toolchain and adds the required dependencies for forge.

Useful links 🔗

• xp.cortexlab.fr
• wiki.cortexlab.fr