Reconfigurable Legged Metamachines that
Run on Autonomous Modular Legs

Chen Yu^{1, *} David Matthews^{1, *} Jingxian Wang^{1, *} Jing Gu¹ Douglas Blackiston^{2, 3} Michael Rubenstein¹ Sam Kriegman¹

¹Center for Robotics and Biosystems, Northwestern University

²Dept. of Biology, Tufts University

³Wyss Institute for Biologically Inspired Engineering, Harvard University

^*Co-first author

TLDR By combining different arrangements of minimal yet athletic agents into a unified body, a diversity of agile legged metamachines were realized.

Overview

Legged machines are becoming increasingly agile and adaptive but they have so far lacked the basic reconfigurability of legged animals, which have been rearranged and reshaped to fill millions of niches. Unlike their biological counterparts, legged machines have largely converged over the past decade to canonical quadrupedal and bipedal architectures that cannot be easily reconfigured to meet new tasks or recover from injury. Here we introduce autonomous modular legs: agile yet minimal, single-degree-of-freedom jointed links that can learn complex dynamic behaviors and may be freely attached to form legged metamachines at the meter scale. This enables rapid repair, redesign, and recombination of highly-dynamic modular agents that move quickly and acrobatically (non-quasistatically) through unstructured environments. Because each module is itself a complete agent, legged metamachines are able to sustain deep structural damage that would completely disable other legged robots. We also show how to encode the vast space of possible body configurations into a compact latent design genome that can be efficiently explored, revealing a wide diversity of novel legged forms.

Metamachines in the wild

Reconfigurable legged metamachines behave autonomously across sand, mud, grass, tree roots, plant litter, mulch, gravel, bricks, concrete, and combinations thereof.

Acrobatic behaviors and resilience to damage

Reconfigurable metamachines behave, resist adversarial perturbations, and adapt to damage. Control policies use internal sensing only; motion capture was not supplied to the policy, it is used only for the behavioral analyses.

Citation

Please use the following BibTeX entry to cite this work:

@article{yu2025reconfigurable,
  title={Reconfigurable Legged Metamachines that Run on Autonomous Modular Legs},
  author={Chen Yu and David Matthews and Jingxian Wang and Jing Gu and Douglas Blackiston and Michael Rubenstein and Sam Kriegman},
  journal={arXiv preprint arXiv:2505.00784},
  year={2025}
  }