World Copy
Enter the lab

Less abstraction.
More interaction.

Look closer. Change a condition. Follow its consequence. Explore the ideas behind physical world simulation.

The model is not
the world.

One workcell, two visual treatments. Move the boundary to inspect how a model and a rendered observation describe the same scene.

WORKCELL / 001
50 / 50
Rendered visual treatment of a humanoid robot workcell.
THE QUESTION

What does the model preserve?
What does reality add?

Follow the signal

One world.
Many ways to see.

Appearance is one channel. Object identities are another. Explore how a shared scene becomes structured information for a robot.

PERCEPTION / 002
50 / 50
NVIDIA Isaac Sim RGB rendering of a robot in a room.
Domain randomized robot training environments.
01 / DOMAIN RANDOMIZATION

Teach what may change.

Vary the surroundings. Keep the task in focus.

RGB, segmentation and bounding box channels for robot training.
02 / SYNTHETIC DATA

Keep the signals together.

Turn one scene into complementary observations.

A world that
pushes back.

Small actions leave a trace. Adjust the force, choose a material response, and place a contact in the field.

EXPERIMENT CONTROLS

01 / MATERIAL RESPONSE
GENTLESTRONG

An illustrative interaction study.
Explore response, not a measured solver output.

CONTACT POSITIONX 0.50 / Y 0.54
FIELD STATUSAWAITING CONTACT
MATERIALWATER / FLUID
APPLIED FORCE0.62 N

TOUCH THE FIELD

WORLD COPY / PROJECT INDEX

A world with
consequences.

CHOOSE YOUR POINT OF ENTRY.
PHYSICAL WORLD SIMULATION FOR ROBOT TRAINING2026

ACKNOWLEDGMENTS

Image credits
& references.

Images are used as visual studies. Depicted organisations and photographers do not endorse World Copy.