Sim-to-real transfer is an important term in the fields of artificial intelligence, automation and robotics. It describes the transfer of skills or solutions that a system - for example a robot - has first learnt in a computer simulation to the real world.
Imagine that a robot has to move goods from one place to another in a factory. Instead of months of training directly on the expensive robot, it is taught everything it needs to know in a safe, virtual environment. There it can practise thousands of times without running the risk of breaking anything. What they have learnt is then transferred to the real robot using sim-to-real transfer.
The big advantage: training in the simulation saves time, money and reduces risks. In addition, situations can be tried out that would be too dangerous in reality. Sim-to-real transfer thus helps to automate processes more quickly and safely test innovative AI solutions before they are used in real life - for example in self-driving cars, industrial robots or drones.
Deutsch 
English (UK)