Computational Ant Lab group at the University of Bristol

Welcome to the homepage of Computational Ant Lab group! Our goal is to use computer science to better understand the emergent group intelligence demonstrated by social insects. In particular, we are concerned with the analysis, simulation and automation of scientific experiments relating to collective decision making in ant colonies.

Biological Focus

Our main focus is on a species of ant intensively studied by our biological colleagues in the Bristol Ant LabTemnothorax albipennis is a species of small European rock-dwelling ant that lives in simple nests within flat narrow rock cavities. Each colony comprises a single relatively large egg-laying queen, about a hundred non-breeding female workers (with no obvious physical specializations among different roles), and eggs of various stage of development. This species is especially amenable to study in the laboratory where whole colonies can easily live in artificial nests made from a couple of microscope slides, and can be induced to emigrate to another nest site by by simply "lifting the lid" off their current home. This affords a unique opportunity to study the complex collective decision making ability of colonies to efficiently choose between alternative nest sites of differing qualities. Of particular interest are the speed-accuracy trade-offs exhibited under varying levels of perceived environmental threat.  At present, the mechanisms by which such ant colonies arrive at a consensus are still poorly understood.

A photograph of Temnothorax albipennis, an example of an artificial nest and an experimental arena involved in such experiments are shown below.

Computational Methods

Our work seeks to combine state-of-the-art methods in computer vision, machine learning and scientific automation to better understand the emergent intelligence of ant colonies. Through a series of final year projects in the Department of Computer Science at the University of Bristol, we have developed the most accurate computational model to date for the behavior of Temnothorax albipennis. We have implemented this model in a cutting-edge gaming engine in order to allow us to conduct "virtual" scientific experiments. This has led to the discovery of new biological hypotheses regarding some seemingly paradoxical behaviors that are seen in real experiments. We are also in the process of developing automated nests that will allow our biological colleagues to conduct revolutionary new experiments on decision making in dynamic environments; and we are gathering finer grained video data on ant behavior. Our work draws heavily on the following computational methods:
  • Computer vision (object tracking and feature analysis)
  • Mathematical modeling (dynamical models and agent-based models)
  • Automated experimentation (3D printing and mechatronics)
  • Machine learning (clustering and equation discovery)
  • Computational Simulation (gaming engines, visualization and interaction)

A robotic arena with two 3D printed nests designed by Cosima Calder (2015) is shown on the left figure below. Interface of SPACE simulator is shown from two different angles on the right figure.

A video of low quorum emigration and a high quorum emigration in SPACE is given below on the left. A comparison between sped-up emigrations in SPACE and during a real experiment is shown on the right.


Project 6.avi

Site created by Gleb Kolpakov
Photograph of an ant colony used for heading logo originally taken by Nigel Franks
Image of a circuit board is in the public domain
Photograph of Temnothorax albipennis is attributed to April Nobile / © / CC-BY-SA-3.0 as required by license

  Detailed Visitor Statistics
web counter