Simulo

Simulo trains robot control policies in simulation. You write the task in Python, preview it, train a policy, then evaluate a checkpoint against the success rule you wrote. Simulo keeps the best and latest checkpoints as a policy you can play, download, keep training, or export to ONNX. The simulo package installs no PyTorch, CUDA, or simulator on your computer; all of that runs in the cloud.

From install to an evaluated policy

pip install simulo
simulo login
simulo create mybot
cd mybot
simulo run preview.py
simulo run train.py
simulo run eval.py --policy policy_<job>:best
simulo policy list

The Quickstart walks through each command and what it prints. Preview before training shows how to fix the task before training, and Evaluate your policy shows how to measure the trained checkpoint and inspect a recorded failure.

Simulo global catalog

The Simulo global catalog is a read-only collection of validated, versioned robots, worlds, and props. You use one by reference from Python, with nothing to download or copy into your app:

cartpole = simulo.Asset.from_registry("simulo/robot/cartpole:v1")

Browse it with simulo asset list --global. The catalog includes:

CartpoleFranka PandaHumanoidUnitree H1JetBotPick-and-placeANYmal CWarehouse worldLab tableMounting stand

Choose and validate assets shows how to inspect a catalog asset and how to publish your own robot to your organization’s catalog.

Explore the docs

Getting StartedInstall the client, sign in, train and evaluate your first policy, and download it.

Core ConceptsApps, jobs, tasks, evaluation, policies and checkpoints, assets, and runtimes.

GuidesStep-by-step instructions for training, exporting, assets, recordings, and live viewing.

ExamplesComplete training apps you can copy and run.

Simulo ConsoleSee your jobs and policies in the browser, read logs, download checkpoints, and cancel jobs.

CLI ReferenceEvery simulo command and flag.

API ReferenceThe public Python API for apps, jobs, tasks, training, and assets.