Harvard engineers 3D-print soft robots that bend, twist, grasp, and move on command
What just happened? When engineers talk about soft robots, precision is usually the missing piece. The challenge has never been in making flexible machines – it's been in making them behave predictably. Now, a team at Harvard says it has solved that problem with a 3D printing technique that encodes movement directly into the material itself.
The method, described in Advanced Materials, replaces the slow, multi-step molding and casting process that traditionally defines soft robotics. Researchers have figured out how to 3D-print structures that twist, curl, or bend exactly as programmed, just by pumping air into their built-in channels.
The work originates from the lab of Jennifer Lewis, a pioneer in multimaterial printing. Graduate student Jackson Wilt and former postdoctoral researcher Natalie Larson integrated several of the lab's existing methods into what they call rotational multimaterial 3D printing, a process that allows multiple materials to flow through a single rotating nozzle.
By continuously spinning this nozzle during printing, the team can determine where each material ends up inside the printed filament – like drawing a helix within a tube. The outer layer, made of a tough polyurethane, forms a durable shell.
Inside sits a gel-like polymer called a poloxamer, familiar from hair care products, which temporarily fills the space where the pneumatic channels will later exist. After the print solidifies, the inner gel is simply washed away, leaving behind meticulously shaped, hollow pathways.
Those channels act as programmable muscles. When pressurized, air or fluid moves through them, forcing the structure to bend, twist, or stretch in predetermined ways. Each filament can contain different orientations or geometries, effectively embedding motion logic inside the material. "We use two materials from a single outlet, which can be rotated to program the direction the robot bends when inflated," Wilt said.
The simplicity of this approach reshapes how soft robots are designed. Instead of fabricating parts separately – casting layers, attaching membranes, and sealing components – the printer renders an entire actuator in one step.
The process requires no hardware rebuild; it's just a matter of adjusting the printing parameters. Complex devices that once took days to assemble can now be redesigned in hours.
To show what this means in practice, the researchers printed two concept pieces: a spiral actuator that unfurls like a flower when inflated, and a gripper with articulated fingers that curl around an object. Both were made in continuous 3D-printed paths.
The implications extend beyond industrial robotics. Such programmable softness could enable surgical tools that adapt to tissue, wearable assistive devices that conform to the human body, or manufacturing grippers able to handle fragile components.
Larson, who has since joined Stanford University as a faculty member, sees this as a conceptual shift for the field. Instead of treating motion as something added to a robot later, she and Wilt argue that the function can now be printed in. The approach effectively makes geometry the code, granting designers direct control over how a soft structure will perform once inflated.