This project provides a framework for applying reinforcement learning (RL) to network simulation research by integrating the OpenAI Gym toolkit with the ns-3 network simulator. It enables researchers to develop and test RL agents for network optimization tasks, offering a standardized interface for RL experimentation within a powerful simulation environment.

How It Works

ns3-gym acts as a bridge between ns-3 simulations and RL agents. It exposes ns-3 simulation parameters and events as Gym environments, allowing RL algorithms to interact with the network simulation by observing states, taking actions, and receiving rewards. The framework leverages ZeroMQ (ZMQ) for inter-process communication between the ns-3 simulator and the Python-based RL agent.

Quick Start & Requirements

• Installation: Requires Linux (Ubuntu 22+ recommended). Install ns-3 dependencies (gcc, g++, python3, python3-pip, cmake), ZMQ, Protocol Buffers, and pkg-config. Download and build ns-3 (version 3.40 or compatible). Clone the ns3-gym repository into ns-3's contrib directory, checkout the correct branch (app-ns-3.36+), and build ns-3. Install ns3gym Python package using pip3 install --user ./model/ns3gym.
• Prerequisites: C++ compiler, Python 3, ns-3, ZMQ, Protocol Buffers, pkg-config.
• Links: ns-3 documentation, ns3-gym GitHub

Highlighted Details

• Enables using any ns-3 simulation script as a Gym environment by implementing a C++ interface.
• Provides examples for cognitive radio channel selection and RL-based TCP congestion control.
• Allows observation of any ns-3 variable or parameter for RL state representation.

Maintenance & Community

• Developed by TU-Berlin (Piotr Gawlowicz, Anatolij Zubow).
• BibTeX citation provided for academic referencing.

Licensing & Compatibility

• The project's license is not explicitly stated in the README. Compatibility for commercial use or closed-source linking is therefore undetermined.

Limitations & Caveats

• The gym.make('ns3-v0') interface may cause errors with newer OpenAI Gym versions; ns3env.Ns3Env() is the recommended alternative.
• The RL-TCP agent is still under development, with existing interfaces for monitoring and controlling TCP parameters.