When you pick up a cup, your fingers adjust a little to its shape. A robot benefits from similar flexibility, especially if its camera estimates the object's position imperfectly. Researchers proposed designing soft grippers^[1] for these small errors rather than assuming that every approach will be perfectly aligned.

A soft gripper is the part of a robot that holds an object while changing shape slightly during contact. The team generated 6,646 candidate designs in a computer model. They considered different objects and grasp positions to look for shapes that would not fail after a small shift in alignment. That makes the design question different from finding the best shape for one perfect attempt^[2].

The researchers then manufactured and physically tested 24 selected designs against alternatives. Real tests matter because a promising simulation does not guarantee that a physical gripper will hold an object. Yet checking 24 examples does not prove that every one of the thousands of computer-generated designs would work equally well.

This approach might help robots sort products or pick up unfamiliar household items. Before relying on it, engineers still need to study material wear, slippery or heavy objects and repeated failures over long periods. The researchers call their design method SimTO-Robust.