Why do robots practise in simulation?
Physical robots can do part of their learning in a virtual environment before trying it for real.
What is it?
Robots can learn much of their movement in a virtual environment, where they can try a million times with no damage and no cost. They then carry the skill into the real world, where friction, light and objects differ from the simulation.
An everyday example
A robot dog learns to climb stairs in simulation before it is tested on real ones.
Why does it matter?
Simulation speeds up robot development and reduces risk while robots learn.
A common misconception
That success in simulation proves safety in reality. Moving from simulation to the real world is one of robotics’ biggest challenges.
In more detail
What is a simulation?
A computational model of the world in which movements can be tested cheaply or more safely.
What is the transfer problem?
In the real world, friction, weight, light and disturbances are different from the model.
Why not practise only in the real world?
Because tests can be slow, dangerous or expensive.
How is the gap narrowed?
Through testing, adjusting the simulation and careful, step-by-step introduction.
Remember: Success in a simulation is not yet proof of safety in reality.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.