SELF-PACED WORKSHOP
How to take this course
Choose observation, fixed-model labs, or training practice. All three follow the same learning sequence.
Choose a learning path
| Path | What you do | What you need |
|---|---|---|
| Observe | Read, inspect the 3D assembly, watch demonstrations, and answer checks. | A modern browser |
| Use fixed models (recommended) | Copy terminal commands, run simulation, and record observations. | A prepared GPU environment |
| Practice training | Add your own training, export, and evaluation. | Additional runtime and disk space |
Start with the demonstrations, then optionally practice training in Chapters 3 and 4. Chapter 6 is an overview; Chapters 7–9 explain frontier methods and integration requirements without new framework installation or large-model training. All reading and browser interactions work without a GPU.
Prerequisites
- Open a Linux terminal, change directories, and copy multiline commands.
- Read basic Python functions, arrays, and parameters. Prior PPO knowledge is not required.
- Understand positions, angles, and velocities; robot coordinate frames are introduced in the course.
A useful lab routine
- Read the expected result before running a command.
- Run commands from the repository root. Use a new output directory for every run.
- Keep the command, seed, and result—including failures.
- Answer the chapter check before moving on.
Reading and exercise durations are planning estimates, not training-time guarantees. The first launch may download assets, compile kernels, and populate rendering caches.