How does an omnidirectional robot with 20 telescoping legs and 20 depth cameras move equally well in every direction without needing to face a particular way?
Photo source:
Argus-robot
For decades, roboticists have built machines by
copying shapes found in nature. Humanoid robots walk on two legs. Quadrupeds
move on four. Insect-inspired designs use six. Each of these approaches
inherits a fundamental limitation from the body plan it imitates: the robot
must face the direction it wants to move. Turning, reorienting, and recovering
from a fall all cost time and energy. Researchers at Duke University's General
Robotics Lab, led by Boyuan Chen, challenged that assumption entirely. Instead
of asking what shape a robot should be, they asked a different question: how
uniformly can a robot accelerate in every direction at once?
That question led to a new design principle
they call dynamic symmetry. Rather than measuring how symmetrical a robot's
body looks, dynamic symmetry measures how equally the robot can move its center
of mass in any direction. The team formalized this concept through a metric
called dynamic isotropy, scored from 0 to 1. Most robots in use today,
including advanced quadrupeds, humanoids, and conventional drones, score below
0.6. The team then simulated more than 1,500 robot configurations to find a
design approaching the theoretical maximum. The result is Argus, named after
the many-eyed giant from Greek mythology, an omnidirectional robot that scores
0.91.
Argus has no front, no back, no top, and no
bottom. It consists of 20 modular, telescoping legs radiating outward from a
central core. Each leg is a single-degree-of-freedom linear actuator mounted at
one of the 20 vertices of a regular dodecahedron, a three-dimensional shape
with 12 pentagonal faces. This geometric arrangement produces a near-perfectly
uniform distribution of instantaneous acceleration capability in every
direction.
Each leg is tipped with a depth-sensing camera,
giving the robot 20 eyes that provide a nearly complete spherical field of
view. This omnidirectional perception remains functional even during aggressive
maneuvers like rolling, tumbling, or rapid reorientation. The combination of
uniform motion capability and all-around sensing means Argus can simultaneously
move, perceive, and interact with its environment without needing to stop and
reorient itself. The modular design also means individual legs can be replaced
independently without disassembling the entire robot.
The broader significance of Argus lies less in
the robot itself and more in the design principle it demonstrates. Dynamic
symmetry provides a general mathematical framework for scoring, comparing, and
designing any robotic system based on uniformity of motion rather than visual
resemblance to biological organisms. The team's simulation sweep of over 1,500
morphologies is publicly available, allowing other research groups to explore
the design space further.
Across those simulations, the team found that
as dynamic symmetry increases toward its theoretical limit, performance
improves consistently across five separate measures: trajectory tracking
accuracy, task success rate, robustness to external disturbances, resilience to
actuator damage, and energy efficiency. These improvements were not
incremental. They became most pronounced as dynamic isotropy approached its
maximum, suggesting that the relationship between symmetry and performance is
nonlinear. The team also explored designs with up to 40 legs that scored even
higher in dynamic isotropy, but these were less practical for a physical
prototype due to the added mechanical complexity.
In physical testing, Argus demonstrated
capabilities across several categories that traditional robot designs handle
separately. It traverses grass, sand, wet ground, and dense forest clutter
without needing to adjust its orientation relative to the terrain. It
self-stabilizes rapidly after being pushed, dropped, or flipped, recovering its
balance without a defined upright position because every orientation is equally
functional.
The omnidirectional robot continues operating
even when as many as three of its 20 legs are disabled. Rather than failing or
limping, the remaining legs redistribute the workload, and the robot's motion
quality degrades gradually rather than collapsing entirely. In simulations, the
team found that this resilience scales directly with dynamic symmetry: the
closer a design scores to 1.0, the more gracefully it handles damage. Beyond
locomotion, Argus can climb walls, carry payloads, and perform whole-body manipulation
tasks where it uses its legs not just for movement but also for interacting
with objects in its environment.
Most commercial and research robots are
designed around specific tasks: walking on flat ground, flying in open air, or
manipulating objects on a table. Argus represents a different approach, one
that optimizes for versatility across tasks rather than performance in any
single one. The ability to move, sense, stabilize, climb, carry, and manipulate
without reorientation makes it relevant to applications where the environment
is unpredictable, such as disaster response, planetary exploration, or
inspection of confined and cluttered spaces.
The research was published in Science Robotics,
and Duke University has filed patent rights for the technology. The simulation
code is available on GitHub under a Creative Commons license. Boyuan Chen, who
directs Duke's General Robotics Lab, has described Argus as an early
demonstration of a wider class of robots built not around biological imitation
but around a fundamental principle of balanced, direction-agnostic performance.
Whether dynamic symmetry becomes a standard design framework in robotics will depend
on how other research groups adopt and extend the principle across different
scales and applications.
Please subscribe to have unlimited access to our innovations.