Harvard's Ant-Inspired Swarm Robots Build and Dismantle Structures

Simple Robots That Collectively Build and Excavate Are Inspired By Ants
Researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) and the Faculty of Arts and Sciences have developed a swarm of simple, ant-like robots capable of autonomously building and dismantling structures without central control. By utilizing minimal physical rules and environmental cues, these robots demonstrate how complex, self-organized tasks can emerge from the interaction between simple agents and their surroundings.

Mimicking Nature Through Stigmergy

The robotic platform, referred to as RAnts, draws inspiration from the biological technique of stigmergy used by social insects like ants and termites. In nature, ants modify their environment and respond to those modifications by emitting pheromones. To replicate this, the Harvard team equipped their robots to respond to "photormones"—light fields that serve as digital equivalents to chemical pheromones.
Each robot senses gradients in the photormone field and leaves signals as it moves, creating a feedback loop that enables coordination across the entire swarm. Operating under a few simple rules, the robots follow these signal gradients, transport building blocks, and deposit materials when specific signal thresholds are met. This process leads to the emergence of nucleation sites, where robots become temporarily confined by the signals they generate, causing structures to form as more agents converge on the location.

Tuning Collective Behavior

A key finding of the study, published in PRX Life, is that the swarm’s behavior can be precisely tuned by adjusting two parameters: cooperation strength, which dictates how strongly robots follow the signal gradient, and the material deposition rate. By modifying these settings, the researchers successfully demonstrated that the same swarm could switch between constructing new structures and dismantling existing ones.
This research introduces the concept of "exbodied intelligence," where collective cognition arises from the ongoing interaction between agents and an evolving environment rather than from the individual agents themselves. According to L. Mahadevan, the Lola England de Valpine Professor of Applied Mathematics, Organismic and Evolutionary Biology, and Physics at SEAS and FAS, this approach shows how simple, local rules lead to robust and adaptive task completion.

Future Applications and Theoretical Framework

The team’s work is supported by a theoretical framework that describes how agent density, communication signals, and environmental structure evolve in tandem. By extending classical biological aggregation theories, the model accounts for environments that change dynamically as agents act upon them.
The potential applications for this technology are broad, ranging from autonomous construction in hazardous environments to planetary exploration and experimental models for studying animal behavior. The study was co-authored by Fabio Giardina and S. Ganga Prasath, with funding provided by the National Science Foundation, the Simons Foundation, and the Henri Seydoux Fund.