Image: Duke University
Most robots are built to look like something. Engineers designing machines to navigate the real world have, for decades, reached for the same reference points: the human skeleton, the dog’s four-legged trot, the insect’s crawl. These biological templates have produced impressive machines, but they carry an embedded assumption that a robot needs a front, a back, and a preferred direction of travel. A team at Duke University’s General Robotics Lab has now challenged that assumption directly, and the result is a machine that looks unlike anything in a robotics catalogue and, more importantly, moves unlike anything that has come before it.
Duke’s omnidirectional robot with no front or back
The robot is named Argus, after the all-seeing giant of Greek mythology, and the name fits. It has 20 modular, telescoping legs radiating outward from a central core, each one tipped with a depth camera, giving it a nearly complete spherical field of view. There is no front, no back, no top, no bottom. It can walk, roll, climb, stabilise and manipulate objects in any direction without needing to turn or reorient itself first. The work, led by engineering professor Boyuan Chen alongside doctoral student Jiaxun Liu and postdoctoral researcher Boxi Xia, has been published in the journal Science Robotics.
The design principle behind Argus and what it actually measures
The conceptual foundation of Argus is a design principle the team developed called dynamic isotropy. Rather than asking what a robot should look like, the principle asks how uniformly it can accelerate in every direction in space. The team quantified this as a score from 0 to 1, where 1 represents a theoretically perfect machine that can push off in any direction with exactly equal force. According to the published study, most advanced robots in use today, including state-of-the-art quadrupeds, humanoid robots, and conventional drones, score below 0.6 on this measure. Argus scores 0.91, approaching the theoretical ceiling. As Chen put it: “When a robot can accelerate equally well in every direction, it stops needing to face the world in any particular way. Forward and backwards become the same. Left and right become the same. The whole problem of robot control changes character.”
Why the dodecahedron geometry of Argus produces near-perfect motion symmetry
Getting to that score of 0.91 required solving a geometry problem first. The team ran more than 1,500 simulated robot configurations to identify which arrangement of legs came closest to their theoretical maximum. The winning design placed 20 identical cable-driven legs at the vertices of a regular dodecahedron, a three-dimensional geometric solid with 12 pentagonal faces. This arrangement produces a near-perfectly uniform distribution of both force and visual coverage in all directions. Each leg is telescoping and cable-driven, meaning it can extend and retract to push against surfaces, and each carries its own depth camera so the robot’s perception matches its physical reach in every direction simultaneously. The result looks less like a machine and more like a sea urchin, which is not coincidental. The study explicitly notes the resemblance, and the geometry behind it is the same principle that gives sea urchins their remarkable mechanical consistency.
Argus navigated forests, sand, and wet surfaces in real-world tests
Building a robot that performs well in simulation is one thing; the Duke team tested Argus extensively in the real world, running it across the Duke campus and surrounding terrain. According to the study, Argus rolled across concrete, grass, dense foliage, soft sand, wet surfaces, and tree bark without losing stability regardless of its orientation. It cleared obstacles up to five inches tall. It climbed vertically between two close parallel walls by alternately bracing and thrusting with different subsets of its legs. It carried a ten-pound payload at near full speed and pushed a large cube around a space while rolling continuously. Doctoral student Jiaxun Liu, co-first author of the paper, said: “The first time we saw it navigate among trees and rough terrain, even under heavy collisions, we knew this was something different.”
How Argus keeps moving even when its legs break or its motors fail
One of the more practically significant findings from the research concerns the robot’s resilience to damage. Because its 20 legs each contribute only a fraction of total locomotion, and because the design distributes force evenly rather than relying on a small number of critical limbs, Argus continues to function even when one or more motors fail, or a leg is broken. This is not a minor advantage. Most robots with fewer limbs face significant degradation in capability or complete failure when a key joint is lost. Argus’s architecture makes it structurally tolerant of partial failure in a way that reflects the same maths that makes it omnidirectional: nothing is so dominant that losing it breaks the system.
The future of robotics beyond biological design templates
The team is explicit that Argus is a proof of concept rather than a finished product, but the implications for robotics design are substantial. Postdoctoral researcher Boxi Xia noted that the robot proves that dynamic symmetry is not just a theoretical exercise; it produces a deployable machine capable of navigating real-world challenges. Chen described Argus as the first member of what he envisages as a broader family of dynamically symmetric machines: “Robots that don’t need to imitate dogs or humans to be agile, tough and useful.” The researchers also modelled designs with up to 40 legs that score even higher on dynamic isotropy, though these remain impractical as prototypes for now, given the added mechanical complexity. The dodecahedral architecture of Argus, however, sits at a useful inflexion point complex enough to approach the theoretical ideal, simple enough to actually build and test in the field. Go to Source
