New AI Gives Soft Robot Hands a Grip That Doesn't Fade

Soft robot grippers are great at picking up delicate items, but they have a hidden weakness. The rubbery materials they're made of slowly relax under pressure, like a rubber band that stretches out. Over time, the grip gets weaker, and the object can slip. This is called stress relaxation, and it has limited how useful these robots are in the real world.

A team of researchers took a new approach. They built a soft gripper that doesn't work blind. It uses a built-in camera to watch how the gripper bends and an infrared sensor to track temperature changes. That data is fed into an AI model that learns how the material is relaxing and predicts what force is needed. If the grip starts to fade, the AI adjusts the gripper in real time — effectively squeezing a bit harder to keep things steady.

In experiments, the system held a constant grip for 280 seconds. The force stayed within an average error of just 0.066 newtons — barely noticeable. That was about 80% more accurate than a fixed gripper and 95% more accurate than one that only checked the grip at the start. The AI didn't just react; it understood how the material would behave over time.

Why does this matter? It's a step toward robot hands that can hold things as reliably as human hands. That could mean safer automation in warehouses, better surgical tools, or more helpful home robots. Instead of dropping your coffee or crushing a strawberry, future robots might know exactly how to hold on — no matter how long they wait.