PhysDBPhysical AI Map

Data

Synthetic data

Generated or simulated observations, trajectories, scenes, and annotations used to supplement real robot data.

data-generationaugmentation

What it is

Synthetic data can come from simulation, world models, or rendering pipelines.

Why it matters

It can expand scenario coverage when real-world data collection is slow, expensive, or risky.

How not to overread it

Synthetic coverage must be audited for domain gaps.

Related edges

generates context for

World foundation models

World model training and evaluation

Synthetic scenes need domain-gap audits.

augments

Robot data

Dataset expansion

Synthetic data must not be counted as real-world proof.