Tutorials¶
A Colab quick start and five tutorials take you from a first network to validation against ns-3. Tutorials 01–05 are Python scripts in tutorials/ at the repository root, and the pages here render the same scripts as executed notebooks, so you can read the outputs without running anything. The quick start is a notebook only, written for Google Colab.
| Tutorial | What you learn | Runs on |
|---|---|---|
00 Quick start in Colab  |
pip install, an L2 engine for 64 × 8 robots, the step dict, a delay CDF and the AoI over time, NRConfig() against a validated preset, strict mode |
free Colab, CPU or GPU, under 3 minutes |
| 01 Your first network | make_engine, submit and step, reading delay, AoI and SINR, partial reset |
CPU, seconds |
| 02 Choosing fidelity | L0, L1, L2-legacy and L2 on the same traffic, the ORACLE and NOCOMM bounds, fitting and loading a surrogate level |
CPU, about a minute |
| 03 Configuring NR | NRConfig, presets, TDD patterns, HARQ processes, downlink, multiple cells, strict mode |
CPU, about a minute |
| 04 Isaac Lab integration | NetModule, MessageHistory, NetEnvMixin, network domain randomization |
CPU for the executed cells; the DirectRLEnv listing needs Isaac Lab 3.0 and is not executed |
| 05 Validating against ns-3 | building and running the lockstep bridge, the Isaac-style ns-3 module, the 5G-LENA sweep replay | needs an ns-3 + 5G-LENA build; not executed here |
The tutorials use tiny sizes (a few environments and robots) so that they finish quickly on a CPU. Their numbers show how to read each output. They are not measurements of the engine or comparisons of the levels.
Running them yourself¶
To regenerate the notebooks rendered here, install jupytext, nbconvert and ipykernel next to the package and run
which writes docs/tutorials/*.ipynb. Tutorial 05 is converted without execution unless NS3BRIDGE_ROOT points to an ns-3 bridge build.
The script does not touch the quick start, which has no script. To re-execute it on a CPU against your checkout: