Process

Sim-to-real transfer

You might say: reality gap

Sim-to-real transfer makes simulated behavior work on hardware despite differences in friction, masses, delays, contacts and sensor noise.

What it does

It determines whether a simulator-trained controller or policy survives physical mismatch.

Ask

Which parameters should be identified, randomized or measured, and over what ranges?

Typical

Domain-randomization studies vary visual and dynamics parameters before transfer to a real robot.

What to look for

1. Simulated behaviour running on hardware
2. Friction, mass, delay, contact or noise mismatch
3. Controller or policy surviving physical mismatch

Connections

• See also: Learned policy

Sources

1. Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Wojciech Zaremba and Pieter Abbeel, Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World, AbstractThis paper explores domain randomization, a simple technique for training models on simulated images that transfer to real images by randomizing rendering in the simulator.
2. Xue Bin Peng, Marcin Andrychowicz, Wojciech Zaremba and Pieter Abbeel, Sim-to-Real Transfer of Robotic Control with Dynamics Randomization, Secs. IV.C and V, randomized dynamics and real-robot transferSimulations are attractive environments for training agents as they provide an abundant source of data and alleviate certain safety concerns.

Use it on your project

teardown-apply: hold "Sim-to-real transfer" (sim-to-real) from the teardown at https://teardownatlas.com/robotics/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal