Definition
Plain language
The hard problem of getting skills a robot learned in simulation to actually work on a real physical robot.
As stated in the literature
The transfer gap between simulated training and real-world deployment in robotics; skills learned in simulation often degrade sharply on hardware, making even modest sim-to-real gains notable.
Also called: sim-to-real gap, sim2real
Why it matters: Closing this gap is what makes cheap, fast simulated training actually pay off on physical robots, so even small gains are meaningful.
For example, a grasping skill that works flawlessly in simulation may fumble the same object once tried on the actual robot arm.
Heard on the show
“But zero to seven, crossing the sim-to-real gap with no retraining at all — that's the part that's actually surprising.”Episode 161 — A Robot That Plays Before You Give It a Job, And Why That Beats Retrying