This Robot Swarm Can Flow Like Liquid and Support a Human’s Weight

Artist's depiction of a swarm of robots forming into a wrench.

With their bright blue bases, yellow gears, and exposed circuit tops, the 3D-printed robots look like a child’s toys. Yet as a roughly two-dozen-member collective, they can flow around obstacles before hardening into weight-bearing tools that push, throw, twist objects like a wrench—and bear up to 150 pounds of weight.

The brainchild of Matthew Devlin, Elliot Hawkes, and colleagues at UC Santa Barbara and TU Dresden, the robots behave like a smart material that shape-shifts into different load-bearing structures as needed. Each smaller in width than a hockey puck, the robots took inspiration from how our cells organize into muscles, skin, and bones—each with vastly different mechanical properties.

Dubbed “programmable matter” and “claytronics,” the concept of robotic materials has long intrigued science fiction writers and scientists alike. Made up of swarms of robots, they can melt and reform, but once locked into a configuration, they have to be stiff and strong enough to hold weight and pack a punch.

“Making this vision a reality would change static objects—with properties set at the time of design—into dynamic matter that could reconfigure into myriad forms with diverse physical properties,” wrote the team.

The new study, published in Science, showcases a proof-of-concept design. Depending on physical and magnetic forces as well as light signals, the robots can form tiny bridges that support weight, collapse into their flow state, and reform as a functional wrench around an object. Each process is controlled by the robot’s integral design.

“We’ve figured out a way for robots to behave more like a material,” said Devlin in a press release.

Unexpected Inspiration

Modular robots and drone collectives have already impressed the robotics community and millions beyond. Over a decade ago, a thousand-bot-strong preprogrammed swarm collaborated with nearby neighbors to self-assemble into complex shapes. While dynamic, they couldn’t support weight. Other designs have been stiffer and stronger but have struggled to reconfigure without breaking group dynamics.

Achieving both properties was “a fundamental challenge to overcome,” wrote the team. For robotic materials to become reality, they need to dynamically shift between a flowing state, in which they can take on new shapes, and a solid state once they reach their final shape.

Nature provides inspiration.

The Power of Three

The team tapped into recent insights gained from the study of embryonic tissues. Starting as a bunch of uniform cells, these tissues can rearrange themselves into multiple shapes and flow to heal tissues. Responding to a bath of biochemical signals within the body, they eventually form a variety of structures—stretchy muscles, stiff bones and teeth, elastic skin, or squishy brains.

“Living embryonic tissues are the ultimate smart materials,” said study author Otger Campàs.

Their versatility relies on three main features.

The first is the force between cells. Imagine being on a completely packed bus. Getting off requires you to push a path across multiple people. Cells are the same. Squishing past each other lets each control where they are in space and time based on their genetic instructions.

The second is coordination. To avoid cellular mayhem, cells use a bunch of biochemical signals to share their positions and movements as they lay out the general landscape of a developing embryo. Finally, cells can grab onto each other—dubbed cellular adhesion—with different levels of strength to build a vast library of tissues with different physical properties.

The robots’ design capture each of these features in 3D-printed hardware.

The bottom of each robot features eight motorized gears dotting the exterior. The bottom isn’t perfectly circular. Some sections are carefully carved out, so that neighbors can always grab onto each other and easily slide off without getting jammed—even when tightly packed. These are a bit like the grooved lids of peanut butter jars. Each gear only slightly peeks out of the housing, enough to grab onto another robot but also easily release it when needed.

Be Part of the Future

100% Free. No Spam. Unsubscribe any time.

To mimic biochemical signals, the team turned to light. Each robot is equipped with light sensors on top and a taped-on polarized film, similar to the material lining some sunglasses. These filters only let light waves vibrating in a particular direction to pass through to the light sensor, telling the robots which way to spin their gears.

Lastly, magnets in small chambers are distributed across the robots’ edges. These can freely roll around and stick to neighboring robots regardless of their position, mimicking cell adhesion.

Robots, Assemble

The team manufactured roughly two dozen battery-powered robots and challenged them to a series of tests. The robots weren’t autonomous: The scientists controlled both the grip strength of the gears and the light signals.

One test started with two towers of robots rotating along each other until they transformed into a rigid bridge. Another began with the robots in a diamond shape that then stretched horizontally into a “mover” that could push a five-pound barbell.

Another test roughly mimicked a workout for your arms. Roughly 20 bots held up two five-pound weights on each side and relaxed only one side when prompted, collapsing into a fluid-like state. All the while, the other side stayed strong.

Even more impressively, the robots swarmed around a nail and solidified to hold it in place. They also hugged a triangle-shaped object in their liquid form and transformed into a wrench capable of twisting the object around. In a demonstration of strength, a collective of 30 robots actively supported a human, weighing roughly 150 pounds, as they stepped across. Then, on command, the structure gradually gave way like mud.

These experiments revealed a surprising quirk. The robots could more easily turn into a fluid-like form when the forces between the robots fluctuated slightly. In contrast, constantly pushing against each other resulted in a deadlock, where no single unit could move, torpedoing the overall dynamics of the robots.

The force fluctuations also saved energy. Returning to the bus analogy, it’s a bit like how wriggling out of a tightly packed human barricade is easier than trying to strong-arm your way through. Adding these fluctuations could be especially beneficial for robots with limited power resources, such as those that run on batteries.

For now, the robot collective has only been tested in about two dozen physical units. But computer simulations of roughly 400 suggest their physical dynamics remain the same and the setup is scalable.

The team is envisioning miniaturizing the system. They’re also eager to explore the technology in soft robots. Like living cells, each unit would be able to stretch and change its shape or size. Although these robots would likely be limited by material properties, a swarm could still significantly change the overall structure and flexibility of any final architecture.

Add in a dose of state-of-the-art control methods—such as AI—to further fine-tune how the units interact and the results could “lead to exciting emergent capabilities in robotic materials,” wrote the authors.

Dr. Shelly Xuelai Fan is a neuroscientist-turned-science-writer. She's fascinated with research about the brain, AI, longevity, biotech, and especially their intersection. As a digital nomad, she enjoys exploring new cultures, local foods, and the great outdoors.

This Autonomous Drone Can Track Humans Through Dense Forests at High Speed

Edd Gent

Jan 31, 2025

MIT’s Latest Bug Robot Is a Super Flyer. It Could One Day Help Bees Pollinate Crops.

Edd Gent

Jan 17, 2025

A ChatGPT Moment Is Coming for Robotics. AI World Models Could Help Make It Happen.

Aaron Frank

Jan 13, 2025

What we’re reading