A 1.8 mm soft robot that grows like a plant root just won a top IEEE robotics award

A 1.8-millimeter soft robot that grows like a plant root, steers with a magnetic field, and senses its own shape in real time just won Best Paper at RoboSoft 2026, the IEEE's flagship conference on soft robotics. The device, described in an IEEE Xplore paper from a University of Leeds and UC San Diego team, is a credible direction for safer minimally invasive surgery. It is also firmly pre-clinical.

"Growing robot" describes a class of soft machines that move more like plant roots than conventional robots. Instead of pushing a rigid body through tissue, they extend new material outward from the tip, everting like a turned-out sock. Most prior designs rely on internal air pressure to drive that extension. The Leeds/UCSD robot is pressure-free: it inverts and pushes a thin silicone tube forward, with no pneumatic chamber behind the tip. That distinction matters inside the body, because an inflated bladder has nowhere to vent if the tip is occluded, and the failure mode is the kind of tissue damage a surgeon would not accept.

The same silicone tube that grows the robot also steers it. Its walls are patterned with magnetized regions the team calls "magnetic DNA," and an external magnetic field bends the tube as it extends. The same walls carry embedded sensors that report the device's shape at up to 500 hertz, so the controller knows the configuration of the entire robot as it moves, not just the position of its tip. Earlier eversion robots typically paired a pressure-driven growing body with a separate steerable tip, or relied on external imaging to track position. This design folds growth, steering, and shape sensing into a single silicone body, rather than stacking them as separate subsystems.

In an ex vivo stomach model, the robot, with a 1.8 mm outer diameter and walls roughly 100 microns thick, demonstrated retroflexion, bending back on itself to reach a target site, and took biopsy samples. That demonstration is the load-bearing result behind the award. It is also the only demonstration reported so far. There is no published in vivo work and no human data. The lead author, Benjamin Calmé of Leeds, frames the work as pre-clinical, with biocompatibility and safety studies still ahead, and any clinical use years away and bound by medical-device regulation.

Calmé trained as a medical doctor in France before moving into engineering and medical robotics, a path he describes as a translation role between surgeons and robotics labs. The design carries that perspective. Pressure-free growth is a safety decision as much as a mechanical one. Sensing the robot's own shape, rather than depending solely on external imaging, is the redundancy a clinician would ask for when the device is operating at the limit of what an endoscope can see. The team's stated next focus is neural and spinal applications, where the concrete task is precise electrode placement, an arena where tip growth and shape feedback may matter more than they do in the gut, where endoscopic visualization is already strong.

The integration has to survive contact with real tissue. Fabrication at 1.8 millimeters is non-trivial: defect rates and unwanted attraction between the inner and outer silicone layers are real engineering problems, not cosmetic ones. The magnetic patterning has to hold up after repeated eversion cycles. The 500-hertz shape feedback has to stay accurate once the device is moving through fluid and against compliant tissue, rather than the open lumen of an ex vivo stomach. None of these disqualify the work, and the award is appropriate for what the team has shown. They are reasons to read the next paper, the one with in vivo data, as the actual test of whether pressure-free, self-sensing growth is a path toward safer surgery or a clever benchtop result.