Robot arrives to eat other robots to grow and heal

Columbia University scientists have created robots that can physically grow, heal and improve themselves by absorbing parts of the environment or other robots, a process called “robot metabolism.”

The research shows that machines start with simple stick-like modules and self-assemble into increasingly complex three-dimensional structures, each transformation that makes them more capable than before.

The study, published in Science Advances, introduces the Truss Link-Magnetic Robot Module, which can expand, shrink and connect various angles to form complex structures. This work represents a fundamental shift to a robot that can operate as an open system that is able to adapt to its physical form rather than keeping the original configuration fixed.

From simple sticks to walking robots

The research team showed how a single truss assembles itself into a two-dimensional shape and then becomes a three-dimensional robot. In a compelling example, a tetrahedral robot integrates an additional link to serve as a cane, increasing its downhill speed by more than 66.5%.

Each truss link measures 28 cm when it shrinks and can extend to 43 cm, which is a 53% expansion rate. These modules have automatic alignment and connection of magnetic connectors, allowing multiple units to be connected from various angles without precise positioning.

“Real autonomy means that robots must not only think for themselves, but also maintain themselves physically.” Philippe Martin Wyder, chief writer and researcher at Columbia Engineering, explained. “Just like biological life absorbs and integrates resources, these robots also use materials from the environment or other robots to grow, adapt and repair.”

Mirror Biology in the Robot Development Stage

Researchers have created a multi-stage development process where robots are gradually becoming more capable. A single link can only crawl forward and backward in one dimension. When the three links are assembled into the triangle, the resulting robot gains two-dimensional navigation capabilities, allowing it to avoid obstacles that are impossible for a single module.

Further assembly creates a “diamond-tail” configuration that is able to overcome the 25mm tall ledge and fold itself into a tetrahedron. The tetrahedral form can be overturned by overturning three dimensions on the obstacle, while the final “ratted tetrahedral” configuration reaches its maximum speed, but with reduced stability.

Key features displayed:

• Self-assembly from single modules to complex 3D structures
• Self-healing by reforming broken connections after impact damage
• Replace the “dead” module by programming component falling off
• Robot assist components, functional units to help others develop
• Gradually improving the transformation of each structure

Bioinspiration drives innovation

The concept of robot metabolism draws inspiration from the way biological organisms use simple components (amino acids) to create complex proteins and entire life forms. Likewise, the truss link system uses standardized modules to generate various functional structures.

“Biologies, by contrast, are all related to adaptation – life forms can grow, heal and adapt,” The famous Hod Lipson is co-author and director of the Creative Machinery Lab. “To a large extent, this capability stems from the modular nature of biology that can use and reuse modules (amino acids) of other life forms. Ultimately, we have to let the robot do the same.”

The research team conducted extensive simulations to quantify the probability of various configurations formed randomly. More than 2,000 simulation runs suggest that some structures (such as tail configurations on diamonds) occur in 44.3% of random attempts, while more complex forms require environmental assistance or operator guidance.

Self-healing and component replacement

This study showed significant self-healing ability. When the robot descends from altitude and breaks the connection, they can independently reform their original shape. The triangle, triangular star and diamond-tail configurations are all successfully recovered from impact damage, thus disconnecting their components.

Perhaps most impressive is that robots can replace the “death” component with programmed cell death similar to biological apoptosis. When the battery of the assembly drops below the critical level, it will automatically disengage and separate from the structure, allowing for integrated functionality replacement.

The researchers also demonstrated a robot-to-robot collaboration, where the built tetrahedral robot helps the flat 2D arrangement transform into a 3D tetrahedral by lifting and positioning components like a crane.

The meaning and application of the future

The technology promises application in disaster recovery and space exploration, where robots must adapt to unforeseen circumstances without human maintenance. The ability to self-heal and reconfigure is critical to long-term autonomous operations in harsh environments.

“Robot metabolism provides a digital interface to the physical world and allows AI to promote not only cognitively but physically, creating a whole new aspect of autonomy,” he said. Wilde explained. “Initially, systems that are capable of robotic metabolism will be used in professional applications such as disaster recovery or space exploration.”

But Lipsen warns that the broader meaning: “The image of self-generated robots reminds people of some harsh sci-fi scenes. But the reality is, when we hand over more and more lives to robots…who will take care of these robots? We can’t rely on humans to maintain these machines. Robots must eventually learn to take care of themselves.”

The research team envisions the future robotic ecology where machines can independently maintain themselves, constantly evolve and adapt to new tasks and environments, a crucial step towards truly autonomous robotic systems that can go beyond their initial programming and design limitations.

Related