Ants make up one of the most diverse groups of animals on the planet, with many species boasting unrivaled physical and social behavior, and an overall species richness that far exceeds that of mammals. Various types of ants can carry up to one hundred times their weight or forage hundreds of meters from their nest, the equivalent of a human carrying thousands of kilograms and walking hundreds of kilometers. These feats have inspired researchers to study not only their anatomy that makes this possible, but how they make use of these capabilities to forage and explore. By recording how individual ants move around a laboratory environment, researchers at the Okinawa Institute of Science and Technology (OIST) have come up with a physics-based model for how these ants navigate, now published in the journal PLOS Computational Biology .
Animal motion has been a topic of fascination for scientists for a long time, though physicists have only really gotten involved in the past few decades. OIST PhD student Jack Featherstone, part of the Nonlinear and Non-equilibrium Physics Unit, explains: “Life is one of the most interesting mysteries out there, rivaling the universal extremes of quantum mechanics or astrophysics in complexity. Physicists have now begun applying many of the quantitative tools developed to study the non-living world to explore why animals move and behave the way they do.”
The foundation of most animal behavior studies begins with observing living specimens and gathering quantitative data about how they behave. The OIST researchers were interested to study one type of ant, known as the long-legged or yellow crazy ant (A. gracilipes), due to its reputation as a particularly aggressive invasive species. As a result, these ants are increasingly colonizing urban habitats as they spread across the globe, making it all the more important to understand how they navigate and explore in artificial environments.
In a clear sign of their widespread presence in Okinawa, the researchers had little difficulty finding specimens. After gathering more than 100 ants from locations like parking lots, walkways, and sitting areas around campus, Featherstone and colleagues placed them alone in arenas in the laboratory to record how they explore the new environment. The researchers emphasize that understanding how these ants move when they are alone is very important, though sometimes overlooked.
“Ants are very famous for their collective interactions, mediated by pheromones or other sophisticated communication systems, but the individual motion of an ant is the building block that makes this collective behavior possible,” says Featherstone.
Following the experiments, the researchers trained a neural network to analyze video recordings of the ants. The network tracked the position of different parts of each ant’s body as it explored, generating spatial trajectories that the researchers could then quantitatively analyze.
The researchers found that the key ingredient to better understand and model the ant behavioral data lay in stochastic, or random, modelling. Famously important to understanding diffusion and the motion of small particles, this approach assumes that the overall, non-random behavior of a system can be constructed from individual, random contributions.
“The behavior of any animal depends on a massive number of variables, from the states of individual neurons in their brains, to the weather around them, to what they had for breakfast. Trying to measure every factor that might be relevant here is impossible; instead, the stochastic approach allows us to explore macroscopic behavior without getting lost in the microscopic details,” says Featherstone.
Their proposed model, which includes a combination of several techniques often used in studying bacterial motion or diffusion processes, can reproduce many of the features of the experimental trajectory data. It also allows them to computationally or mathematically derive predictions about how the ants might behave in new environments, which they hope to use to study how these ants interact with other species, either as predators or as prey.
As one of the most rigorous descriptions of ant exploration behavior — including the ability to simulate ant-like motion — this study may help other researchers understand how locomotion fits into the rest of an ant’s life; external stimuli, in the form of food, water, predators, or obstacles will have the ants adjust their behavior from the baseline that this work provides. In the future, it could even be used to better understand and contain invasive ants. And even beyond ants, this modeling approach could be easily adapted to explore how navigation, foraging, and exploration vary throughout the rest of the animal kingdom.
PLOS Computational Biology
Computational simulation/modeling
Animals
Stochastic modeling of long-legged ant A. gracilipes locomotion in laboratory experiments
6-Aug-2026