A fresh study reports that straightforward ant-inspired robots can assemble clusters and tunnel through material without a boss, a map, or any step-by-step blueprint.
The findings suggest a route to machine teams that can continue operating in chaotic environments where centralised control and communications tend to fail.
Swarms operate independently
In a compact test arena, RAnts - small wheeled robots inspired by ants - moved plastic blocks while responding only to the light projected beneath them.
The work was directed by Professor L. Mahadevan at Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS).
By recording the swarm’s light footprints, Mahadevan’s group demonstrated that purely local signals can trigger both building up and breaking down.
Those same signals were then used to swap out top-down instructions for feedback-driven behaviour.
Ant behavior shapes design
Ants and termites frequently coordinate by altering the spaces around them, including termite mounds: tall nest structures that can ventilate colonies as daily temperatures rise and fall.
Biologists refer to this as stigmergy - indirect coordination via modifications left in a shared environment - where small individual actions become guidance for the group.
In insects, pheromones - signalling chemicals left behind - can label routes that other workers then follow or strengthen.
For these robots, projected light filled the same role, meaning a trail could steer later movement without speech or radio links.
Signals guide the swarm
Rather than relying on scent, the robots tracked photormones - projected light trails that act as stand-ins for pheromones.
Each robot compared brighter and dimmer light using two underside sensors, then steered towards the stronger trace as it travelled.
With every circuit, the robots added additional light to the floor, while the signal gradually decayed once they moved on.
Newer trails drew in more robots, and the fading of older trails prevented past activity from dominating the arena indefinitely.
Clusters create work areas
As light trails were laid down repeatedly, they formed nucleation sites - initial hotspots where building first concentrates - when multiple robots looped around the same patch.
This feedback produced a trapping instability: a self-reinforcing loop that holds robots in place for a time because their own trace keeps luring them back.
In trials, the robots were tuned so that a trap typically required five or more robots, reducing the chance that a single machine could take over the process.
“Our new study shows how simple, local rules can lead to the emergence of complex task completion that is self-organized and thus robust and adaptive,” said Mahadevan.
Simple rules shift tasks
After traps appeared, the swarm could be pushed into different behaviours when researchers changed just two parameters in the control rules.
Cooperation strength - how strongly robots bias their motion towards the brighter parts of a trail - drew them into shared work areas when set to a high value.
Deposition rate - the speed at which robots add or remove material - determined whether those work areas accumulated blocks or were stripped back.
When both parameters sat within the appropriate window, the very same simple rules produced distinct tasks rather than locking the system into one routine.
Building and digging emerge
Adjusting the deposition setting alone was enough to switch the task from construction to excavation, without introducing any supervisor.
When the setting favoured depositing material, the robots collected blocks into orderly piles where the light signal intensified.
When the setting favoured removal, the machines instead carried material away, creating openings and routes through an existing structure.
Earlier studies of ant excavation and robot digging illustrated how simple animals and machines can excavate, but this approach supported both directions of change.
Theory explains the swarm
Beyond the physical arena, the SEAS team developed a continuum model: equations that treat the group as flowing densities.
Rather than simulating every wheel movement, the model tracked robot density, light signals, and material as they evolved across space.
This modelling choice connected individual rules to collective patterns, including dispersed heaps, tight clusters, and advancing excavation fronts.
A phase space - a chart of possible group behaviours - allowed researchers to select building or digging without rewriting the robots’ rules.
Practical value emerges
Centralised control can break down when machines enter collapsed buildings, disaster areas, or remote planets where communication is slow.
Local rules provide an alternative, because each robot needs only nearby signals and a means to transport material.
The team described the approach as exbodied intelligence - coordination emerging from workers plus surroundings - since the workspace stores a short-lived record of useful recent actions.
Potential applications include hazardous construction and planetary exploration, although real-world swarms will still require more capable hardware and robust safety constraints.
Gaps to guide research
The current demonstrations relied on simple blocks, projected light, and a controlled floor, rather than a genuine, open construction environment.
Because the arena was constrained, the robots were not choosing which end structure would be most effective for a practical goal.
Future versions may need outcome selection - rules that promote useful structures, not merely any structure that can form.
Even so, the results give engineers a clearer, lower-cost way to probe cooperation before deploying more expensive machines.
Natural teamwork emerges
Across both robots and insects, the central message is clear: groups can organise when their actions leave traces that others can detect and respond to.
That evidence may help researchers compare insects and machines while staying realistic about the limits - simple rules still leave plenty of hard design work to do.
Comments
No comments yet. Be the first to comment!
Leave a Comment