The researchers showed how individual Truss Links self-assembled into two-dimensional shapes that then could morph into three-dimensional robots. These robots then further improved themselves by integrating new parts, effectively "growing" into more capable machines. For example, a 3D tetrahedron shaped robot integrated an additional link that it could use like a walking stick to increase its downhill speed by more than 66.5%. 

"Robot minds have moved forward by leaps and bounds in the past decade through machine learning, but robot bodies are still monolithic, unadaptive, and unrecyclable,” says Hod Lipson, co-author and James and Sally Scapa Professor of Innovation and chair of the Department of Mechanical Engineering at Columbia University, and director of the Creative Machines lab where the work was done. “Biological bodies, in contrast, are all about adaptation - lifeforms, can grow, heal, and adapt. In large part, this ability stems from the modular nature of biology that can use and reuse modules (amino acids) from other lifeforms. Ultimately, we’ll have to get robots to do the same - to learn to use and reuse parts from other robots. You can think of this nascent field as a form of ‘machine metabolism.’”

Researchers envision future robot ecologies where machines independently maintain themselves, growing and adapting to unforeseen tasks and environments. By imitating nature's approach—building complex structures from simple building blocks—robot metabolism paves the way for autonomous robots capable of physical development and long-term resilience.

"Robot Metabolism provides a digital interface to the physical world and allows AI to not only advance cognitively, but physically—creating an entirely new dimension of autonomy," says Wyder. "Initially, systems capable of Robot Metabolism will be used in specialized applications such as disaster recovery or space exploration. Ultimately, it opens up the potential for a world where AI can build physical structures or robots just as it today writes or rearranges the words in your email."

Lipson concludes with caution: “The image of self-reproducing robots conjures some bad sci-fi scenarios. But the reality is that as we hand off more and more of our lives to robots - from driverless cars to automated manufacturing, and even defense and space exploration. Who is going to take care of these robots? We can’t rely on humans to maintain these machines. Robots must ultimately learn to take care of themselves.”

This work was supported by the NSF AI Institute in Dynamic Systems (NSF NIAIR COLUM 5260404, SPONS GG017178), NSF NRI (NSF NRI COLUM 5216104, SPONS GG015647), and DARPA TRADES (DARPA TRADES COLUM 5216104, SPONS GG012620).

Lead Photo Caption: These Truss Links self-assembled to form a tetrahedron.
Lead Photo Credit: Creative Machines Lab

About the Study

Title: “Robot Metabolism: Towards machines that can grow by consuming other machines”

Authors: Philippe Martin Wyder, Riyaan Bakhda, Meiqi Zhao, Quinn A. Booth, Matthew E. Modi, Andrew Song, Simon Kang, Jiahao Wu, Priya Patel, Robert T. Kasumi, David Yi, Nihar Niraj Garg, Pranav Jhunjhunwala, Siddharth Bhutoria, Evan H. Tong, Yuhang Hu, Judah Goldfeder, Omer Mustel, Donghan Kim, and Hod Lipson

For interview/questions contact: Philippe Wyder: [email protected] +1 (347) 604 4450

Hod Lipson: [email protected] +1 (607) 592 4383 

COI: The authors declare no financial or other conflicts of interest.