A robot may recognize an object correctly yet place its gripper slightly off target. Rigid fingers can lose contact, while a soft gripper may deform around the object. Researchers led by Kurt Enkera introduced SimTO-Robust, a computer-aided design method intended to make grippers tolerant of such errors through their physical shape.

Instead of optimizing for one ideal object and pose, the method simulates contact with several geometries and grasp positions. The resulting forces guide improvements to the compliant structure. Part of the robustness problem therefore moves from complicated control software into mechanical design. Cameras still matter, but small positioning errors may become less damaging.

The team generated 6,646 candidate designs and physically evaluated 24. Comparisons included ten specialized soft designs, two rigid baseline grippers and one conventional general-purpose soft gripper. In total, 37 devices were tested with five objects featuring distinctive shapes, six grasp poses and progressively stronger external disturbances during pick-and-place operations.

On average, SimTO-Robust designs transferred objects more reliably than specialized soft designs and rigid baselines. The strongest candidates tolerated larger disturbances than the other tested devices. These findings apply to the reported materials, objects and conditions, not to every possible gripper or load.

The broader lesson is that useful mechanical design can make a robot forgiving before the controller intervenes. That may matter in sorting, assembly and handling irregular objects. Long-term wear, contamination, cycle speed and performance with heavier or unfamiliar items remain open questions for practical deployment.

An implementation team should define acceptance criteria in advance: results on representative workloads, recovery time after failures, operator intervention and support cost. A demonstration of feasibility is not a substitute for measuring these outcomes at a customer site.