Symmetry is everywhere in nature, from the bilateral form of vertebrates to the radial geometry of starfish. For decades, roboticists have tried to copy these shapes and their abilities with bodies that look like humans, dogs or insects.
Now, roboticists at Duke University argue that what really matters isn't how a robot looks, but how uniformly it can act in any dimension in space. Guided by this concept, the team simulated more than 1,500 robot configurations to arrive at a design approaching their theoretical maximum.
The result is Argus: a robot with no front, no back and 20 modular, telescoping legs radiating out from the central core, each tipped with a depth camera. With a resemblance closer to a sea urchin than any commercial robot, the design proved robust.
Argus easily traverses forests, wet surfaces and sand; self-stabilizes quickly after being pushed; reorients instantly in any direction; climbs vertically between close walls; and even carries and pushes payloads around a given space.
The results appear online May 27 in the journal Science Robotics.
"Watching Argus move is unlike watching any other robot we've worked with," said Jiaxun Liu, co-first author and Ph.D. student in Duke's General Robotics Lab. "The first time we saw it navigate among trees and rough terrain, even under heavy collisions, we knew this was something different."
At the heart of Argus's unique abilities is a new mathematically derived design principal the team has dubbed dynamic isotropy. The term scores from 0 to 1 based on how uniformly the robot can accelerate its center of mass in every direction. Most robots in use today, including state-of-the-art quadrupeds, humanoids and conventional drones, score below 0.6. Argus scores 0.91, approaching the theoretical maximum.
In their simulations to arrive at Argus's design, the team showed that as dynamic symmetry rises toward its theoretical limit, performance improves across nearly every measure that matters in robotics. This includes trajectory tracking, robustness, energy efficiency, resilience to damage and success on difficult terrain. The principle works as a unified yardstick that can be applied to existing platforms.
"Most robotics research has framed symmetry as a question about the body, but we argue that the more powerful symmetry is at the level of what the robot can do," said Boyuan Chen, who leads this research and directs Duke's General Robotics Lab. "When a robot can accelerate equally well in every direction, it stops needing to face the world in any particular way. Forward and backward become the same. Left and right become the same. The whole problem of robot control changes character."
Named after the all-seeing sentinel of Greek mythology, Argus pairs whole-body actuation with whole-body perception. Each of its 20 modular, telescoping, camera-equipped legs are arranged at the vertices of a regular dodecahedron—a 3D shape with 12 pentagon faces. This setup produces a near-perfectly uniform distribution of instantaneous acceleration and a near-perfectly uniform field of view in every direction.
What this design enables is striking. In experiments on the Duke campus, on sand and in forest trails, Argus has demonstrated a wide range of abilities learned from simulations alone:
- Rolling across concrete, grass, dense foliage, soft sand, wet surfaces and bark regardless of its orientation, including obstacles up to five inches tall;
- Stabilizing itself rapidly after being pushed;
- Adapting to damage by continuing to move even when three of its legs are broken;
- Carrying a 10-pound payload at nearly full speed;
- Climbing parallel vertical walls by alternately bracing and thrusting with different subsets of legs;
- Tracking and pushing a three-foot cube while continuously rolling.
These successes extend well beyond Argus, which was built to explore and learn from these design concepts just as much as to demonstrate them. Dynamic symmetry, the team emphasizes, is a general framework providing a way to mathematically score and compare a wide range of robots and to design new ones from scratch. The 1,500-morphology simulation sweep released with the paper allows other research groups to explore the design space directly.
"Argus is an existence proof," said Boxi Xia, co-first author and postdoctoral researcher at Duke's General Robotics Lab. "It shows that designing for dynamic symmetry isn't just a theoretical curiosity. It produces a robot you can deploy in the wild, on uneven ground and in clutter, even in low-gravity settings. It changes what's possible."
"We see Argus as the first member of a much broader family of dynamically symmetric machines," Chen added. "Robots that don't need to imitate dogs or humans to be agile, tough and useful. Robots designed from a deeper principle; the same principal nature uses to build everything from viruses to starfish."
This work is part of the long-term mission at Chen's General Robotics Lab to create Discovery Robotics, building machines that learn, act and collaborate by discovering how the world works. Where most robotics research designs robots to execute predefined tasks, Chen's group designs robots as instruments of discovery, capable of revealing new principles about themselves, their environments and the physics that govern both. Argus is a direct expression of that mission: a robot whose body was effectively searched for and discovered across more than 1,500 candidate designs against a first principle of design theory.
"We don't just want robots that follow instructions," Chen said. "We want robots that help us learn things about the world we couldn't learn any other way, and that sometimes means discovering the right body for the question first." More about the lab's research agenda is available at https://generalroboticslab.com .
