Researchers at the University of Amsterdam have created metamaterials that can learn to change shape and behave in ways similar to living systems. Unlike normal materials that always respond the same way to force, or robots that follow fixed programs, these metamaterials adapt without any central control unit like a brain.
The worm-like structures are made of identical motorized hinges connected by an elastic skeleton. Each hinge contains a small microcontroller, a tiny computer that measures rotation, remembers past movements, and shares information with neighboring hinges. The hinges can apply torque, a twisting force, to adjust their stiffness and preferred position. Through repeated training with examples, the material learns to adopt new shapes. It can forget old shapes, remember several shapes at once, and switch between them. This allows tasks such as gripping objects or moving across surfaces, known as locomotion.
Advancing toward adaptive materials
The learning process uses a method where the material is shown desired shapes and gradually updates its internal properties to match them. Once trained, the metamaterial responds automatically to inputs by changing shape. Researchers observed that learning enables the system to evolve, opening many future possibilities. Earlier work from the same lab produced objects that could roll or crawl on uneven ground, but could not learn new behaviors.
Future plans include teaching the materials time-dependent actions, such as different crawling or rolling movements based on surroundings. Studies may also explore learning under uncertain conditions with noise, which could make the systems more flexible and robust. Interest in such adaptive materials has grown, and a Dutch research program now supports work on materials that learn and how to use them responsibly.
The metamaterials were described in a paper published in Nature Physics. The approach points to new possibilities in robotics and smart materials that can adjust to changing environments without complex programming. All processing occurs locally within the structure itself.