Squishy Robot Arms Can Now Bend Precisely in Tight Spaces

Most robots are stiff and predictable, but 'soft' robots are flexible, like an elephant's trunk or an octopus tentacle. That flexibility lets them squeeze into tight spaces and work safely near people. But it also makes them very hard to control: bend one part, and the whole arm may sag or wobble. Previous controllers only aimed at where the tip should go. For real tasks inside pipes or alongside humans, you need the whole arm to hold a specific shape, especially when it's carrying something or being pushed around.

Researchers at Arizona State University and collaborators built a new control system that gets around this problem. Instead of trying to compute every bend from scratch, they use a mathematical trick that summarizes the arm's behavior with simple 'observables' — like checking the wrist while also watching the elbow. Their system combines this knowledge with a method of planning ahead, so the robot can react in real time, like a driver steering a long, wobbly bus.

In tests, the controller handled soft arms with three and five segments — segments are like individual bendy sections — at speeds up to about 1.3 miles per hour, which is fast for a squishy robot. It kept tracking accurately while carrying a payload as heavy as 400 grams (about the weight of a can of soup) and recovered quickly after being bumped with a serious sideways force. In a simulation, it managed a robot arm with 10 independently controlled segments, showing the approach can scale up.

The team also ran a 'confined-space' demonstration threading the arm through a narrow opening, pointing toward future inspections inside engines or pipes. The bigger takeaway: this research helps close the gap between bendy laboratory robots and practical robots that can safely operate in the messy, cramped, unpredictable spaces where people live and work.