Today's robots often stay the same once built. They are closed systems, meaning they cannot add to themselves, fix damage on their own, or change to fit new places. Researchers at Columbia University have now made robots that can do these things. They use materials from around them or from other robots to grow stronger, repair breaks, and get better at tasks.
The new idea is called robot metabolism. This is a process where machines take in and reuse parts, much like how living things eat and use food to grow or heal. The study, published in Science Advances, shows how this works. The researcher argue that true freedom for robots means they must think for themselves and also keep their bodies going, just as life forms take in resources to adapt.
The robots are based on truss links. These are simple bar-shaped pieces with magnetic ends that can stretch, shrink, and join at different angles to make bigger shapes.
Demonstrating robot metabolism
In tests, single truss links joined on their own to form flat two-dimensional shapes. These then turned into three-dimensional robots. The robots improved by adding new parts, like growing. For example, one robot shaped like a tetrahedron, which is a pyramid with four triangle sides, added an extra link. It used this like a walking stick to move downhill over 66 percent faster.
While robot thinking has advanced through machine learning, robot bodies remain monolithic. In contrast, biological bodies adapt because they are modular, built from reusable parts like amino acids, the building blocks of proteins. With machine metabolism, robots could learn to use parts from others.
The researchers picture future groups of robots that care for themselves without help. By copying nature, using basic blocks to build complex forms, this could lead to machines that handle unknown jobs in tough spots like disaster areas or space. Robot metabolism could allow artificial intelligence (AI) to grow bodies as well as minds.