A project undertaken at Macquarie University, and supervised by Dr Chris Reid.
Ants are masters of collective behaviour, where coordination and communication are cornerstones of nearly every aspect of their daily lives. Some ants, such as the weaver ant (Oecophylla smaragdina) take cooperation to the extreme, forming self-assembled structures such as bridges, hanging chains and pulling chains to perform feats beyond the capabilities of any single ant. Ant self-assemblages are formed using simple agents and interaction rules, yet the emergent structures are sophisticated, forming when and where required, adapting to environmental conditions, and self-repairing when damaged. This project aimed to understand the process of self-assembly by;
i) inducing colonies to self-assemble pulling chains, bridges and hanging chains in the laboratory.
ii) Individually and uniquely marking and tracking individuals using state-of-the-art technology to provide the first comprehensive quantification of behaviours during self-assembly.
iii) using this data to statistically link individual-level behaviours to group-level outcomes, forming a modelling framework that could benefit many areas, including health (models of embryonic development, wound healing etc) and engineering (swarm robotics control).
We developed procedures and methods for long-term rearing and maintenance of weaver ant colonies in the lab, a feat which has been rarely attempted, as these are considered to be a highly difficult species of ant to maintain under laboratory conditions. We successfully developed experimental protocols for the repeatable and reliable production of bridges, hanging chains and pulling chains under laboratory conditions, in scenarios where both group level and individual level behaviour can be recorded, observed and decoded.
An extensive analysis of hanging chain formation by our PhD student Daniele Carlesso has revealed several new aspects to this unique collective behaviour. The majority of ants join only at the tip of the chain, and the probability to join at the tip is negatively correlated with the distance to a visual target platform below. Ants maintain a constant probability of joining that is not influenced by the current state of the chain or traffic conditions, yet they do modulate their leaving probability based on several factors including the distance to the visual target and the rate of traffic flowing in the upwards direction along the chain. An analytical model based on our observations shows strong support that the simple, local level behaviour of modifying the probability to leave based on visual and traffic-flow information is a sufficient and robust method to generate a functional self-assembled structure. This represents the first piece of the overarching statistical framework linking individual-level behaviours to group-level outcomes. This part of the project has led to new collaborations where we aim to develop an agent-based model to gain further insights into the system, and drive new experimental questions.
Research assistant Justin McNab has completed an investigation of the tensile forces inside weaver ant bridges of different component lengths across a gap. By inducing the ants to build a bridge across an otherwise impassable gap, with one end of the platform connected to a load cell that measures the pulling force of the bridge, we found that weaver ant structures are far more fluid than was previously thought. In contrast to army ant bridges, where individuals remain linked to each other with a high tensile force across strong bonds between their legs, weaver ant bridges displayed a much higher rate of turnover, and a much less consistent tensile force across the structure, regardless of bridge length or the number individuals in the structure. Masters student Madelyne Stewardson is now completing a similar study on pulling chains that specifically answers the question; how does the number of ants in a pulling chain affect the overall force output of the chain? This question has implications for the design of future cooperative swarm robots, as well as driving new insights into the mechanistic underpinnings of contractile units, for example sarcomeres in muscle tissue.
We developed a novel tagging and tracking method specifically for weaver ants, that involved the production of unique, QR code-like 1.5mm diameter tags that are glued onto the thorax of each ant, that assist a computer vision system in tracking each ant’s ID and location during behavioural experiments. We are just beginning to fully utilise this new tool in understanding individual-level behaviour during the production of sophisticated group-level colony outcomes, with progress slightly hindered by COVID-19. As a bonus, an unexpected outcome of this approach was the development of a new tracking system that doesn’t require the use of tags, which will be a major boon to behaviour labs worldwide. Developing this system required the use of tagged ants for verification and training of the machine learning algorithms for the tagless method, and we will soon publish the method, and utilise it to further our understanding of collective behaviour.
One overarching aim of the project was to assist in practical applications of cooperative behaviour and self-assembly, one of which is in swarm robotics control. We have begun to develop prototype swarm robots that use behavioural algorithms directly inspired by the ant self-assembly experiments to allow them to join together to form functional structures in the absence of any direct control or pre-planned blueprints. This application will be especially useful in scenarios where pre-planning and direct human control are impossible, such as in search-and-rescue operations and habitat building and exploration in extra-terrestrial environments. Outputs from the Hermon Slade project have directly contributed to successful seed-funding for this aspect of the project.