IC Biovision

Social insects

Ant foraging networks are used to transport resources and/or information during foraging, and also for exploration, emigration and for coordinating colony defence. Just as the functioning and success of modern cities is dependent on an efficient transportation system, the effective management of traffic is essential to insect societies. Furthermore, networks are ubiquitous in nature, and the efficiency of such networks may determine the fundamental scaling properties of certain organisms. The foraging networks of ants provide unrivalled opportunities to quantify both the behaviour of individual items of traffic and larger-scale patterns of traffic flow. For this reason they are ideal subjects to test mathematical models that link the behaviour of small components (here, individual ants) to the overall efficiency of the dynamic structures they generate (see image from an individual-based simulation of trail formation, left-bottom).

I have been studying the organisation of traffic flow in army ant colonies (Eciton burchelli) in Panama (top image). Colonies of this species stage huge swarm raids with over 200, 000 foragers that transport more than 3, 000 prey items an hour over raiding columns that can extend over 100 metres. Yet individuals are small (mostly less than 1cm long) and almost completely blind. Using image analysis to track the movement of ants on trails, and individual-based computer models of pheromone trail-following (osmotropotaxis), I am investigating how individual behaviours lead to crucial properties including information on directionality, lane formation and the minimisation of congestion.

Published in Proceedings of the Royal Society of London, Series B. See the abstract.

Media reports and interviews

Editor's choice, "Avoiding Gridlock", Science 299, 19

News focus, "Getting the behavior of social insects to compute", Science 295, 2357 article.

"The ants go marching -- and manage to avoid traffic jams", Princeton Weekly Bulletin article.

"Army ants march to military efficiency", New Scientist.

"Army ants obey traffic plan to avoid jams", National Geographic article.

"Ant traffic flow: raiding swarms with few rules avoid gridlock", Science News article.

"Ants offer lessons in urban living", The Guardian

"An urge to organise", The Philadelphia Inquirer.

"Of ants and men: traffic flow", The Today Programme, BBC Radio 4.

"Natural technology", Nature, BBC Radio 4.

"Future watch: ant traffic", Radio South Africa.

Other properties of social insect colonies I am interested in are activity cycles exhibited by certain species. Using Computer Vision approaches we were able to reveal the spatial properties of these waves for the first time, and developed coupled-oscillator models to explain the regulation of such waves within colonies. See the abstract, or download the paper (Proc. R. Soc. Lond. B).

I am continuing investigations of ant colonies using Computer Vision (see below).

Collective motion

The integration of individual behaviours within fish schools leads to a synchrony of motion that is captivating (top image on the left). Similar patterns can be seen in bird flocks, where the volume and shape of the group change as it arcs overhead, yet the aggregate can remain cohesive (middle image).

I have been developing individual based models of animal grouping in three-dimensional space (bottom image) to investigate how individual behaviours result in collective patterns (including loose 'swarms', collective directional motion and the generation of a 'torus' formation where individuals perpetually rotate around an empty core). This modelling has been useful in revealing theoretical properties of animal groups, including "collective memory", where the previous history of the group structure influences the collective behaviour exhibited as individual interactions change, even though individuals have no knowledge of what that history is. I have also been exploring how behavioural differences among individuals influence the internal structure of such groups, and how individuals using simple, and local, rules can change their spatial position within a group (e.g. to move to the centre, the front, or the periphery) in the absence of their current position within the group as a whole.

With Dr. Jens Krause and the fish behaviour research group at the University of Leeds I am working on models to better understand group choice behaviour in fish. We have also developed new techniques to study spatial positions of individuals within fish schools in the field as well as imaging techniques to automatically track the motion, and record interactions among, individuals in the laboratory (see Computer Vision, below).

Current papers on this topic include:

Effective leaderhsip and decision making in animal groups on the move, Nature

Modelling density-dependent fish shoal distribution in the laboratory and field, Oikos: see abstract.

Context-dependent group choice in fish, Animal Behaviour: see abstract.

Mechanisms underlying shoal composition in the Trinidadian guppy, Oikos: see abstract.

Self-organization and collective behavior in vertebrates, Advances in the Study of Behavior: see abstract.

When fish schools meet: outcomes for evolution and fisheries, Fish and Fisheries: see abstract.

Collective memory and spatial sorting in animal groups, Journal of Theoretical Biology: see abstract, or download paper.

Effects of parasitism and body length of positioning within wild fish schools, Journal of Animal Ecology: see abstract, or download paper.

A grid-net technique for the analysis of fish positions in free-ranging fish schools, Journal of Fish Biology

Social organisation of fish schools, Advances in Ethology.

Also see Information transfer / social learning below.

Tropical cats / Automated telemetry

Collaborators: Dr. Martin Wikelski (Princeton University), Dr. Roland Kays (New York State Museum), William Cochran (Champaign, IL), Jamie Mandel (STRI) and others.

Starting in July 2002, I have been working on the Automated Telemetry project on Barro Colorado Island (BCI), in the Republic of Panama (a site of the Smithsonian Tropical Research Institute, STRI). One of the main problems associated with studying animal behaviour is accurately recording the motion of individuals, interactions among individuals and interactions between individuals and their environment. I have been working extensively on digital computer imaging techniques to facilitate tracking of organisms visually (see Computer vision, below). However, some of the most difficult animals to study are those that travel over large areas and/or move through complex habitats, where computer vision techniques cannot be applied. Consequently there is very little quantitative information about the movement patterns and behaviours of large animals within the tropical rainforest. The result of this dearth of information is that we have little quantitative data with which to address some of the most fundamental questions in biology, ecology and conservation, including: species interactions and the evolution of diversity, competition, predation, seed dispersal, the effects of habitat fragmentation and human disturbance.

In this pioneering project, we aim to set up an automated telemetry system (ATS) that will facilitate the automatic radio-tracking of tagged animals within a rainforest. Wireless communication between 7 canopy towers (installed for this project), and the laboratory, will mean that the positions of animals will be available digitally in real-time, online, allowing registered users access to the trajectories of individual animals on Barro Colorado Island wherever they are in the world. I am also developing a 3-dimensional virtual landscape of BCI to allow visualisation of our data, triangulation accuracy studies and an environment for computer models of animal movement in the rainforest (see image to the left, bottom).

One of the initial projects we have begun, to test this system, is a study of tropical cats, including the ocelot (pictured to the left, top). Further projects will involve large frugivorous birds (such as toucans) and high-flying bats.

For more information on this project, and the latest developments, visit "The Automated Telemetry Project" website.

Locust swarms

Collaborators: Prof. Steve Simpson, Prof. Phillip Maini, Dr. Dave Sumpter (University of Oxford)

"Understanding group formation and the collective movement of locusts" (Funded by the EPSRC)

Despite the importance of understanding group dynamics to ecological processes, many phenomena of collective behaviour are still only qualitatively understood. Here we propose to address this topic by using a combined empirical and mathematical modelling approach to study the collective behaviour of the desert locust, Schistocerca gregaria (Orthoptera, Acrididae). This is a particularly good model organism for our study since much is already known about the physiology and sensory system of this species, it is amenable to experimental manipulation, and yet very little is known about how such properties translate to the large-scale mobile groups of this organism. The desert locust is notorious for forming large swarms, and plagues of this species travelling across Africa and Asia can cover over 20% of the land surface of the Earth, where the impact on agriculture, and hence quality of human life, can be devastating.

When they aggregate and gregarise, flightless juvenile locust nymphs, known as “hoppers”, exhibit a distinctive pattern of behaviour known as marching: they form large bands (as opposed to adult flying ‘swarms’) that move between food patches (see image to the left, top). A band is usually a broad phalanx of individuals, and can vary in size from several thousand, to hundreds of thousands of individuals, thus representing some of the largest animal groups found in nature. The largest bands may be tens of kilometres long, and yet they remain cohesive, travelling up to 2km per day. As bands move they tend to merge with others. Where conditions are sufficient for the persistence and growth of bands, locusts are prone to swarm formation. In contrast, in environments in which aggregations persist with less ease swarm formation is inhibited. It is of great practical interest, therefore, to predict the behaviour and movement of hopper bands in their natural environment. A great deal is known about the behaviour and time-budgets of individual locusts in the laboratory, but this does not explain the patterns of activity observed in the field. In particular, gregarious locusts on their own do not march but spend most of their time quiescent. There exists a scattered and often anecdotal literature on the behaviour of marching hopper bands, as well as some laboratory work. What is required is an analytical treatment that has the power to explain and predict the behaviour of hopper bands.

This study will investigate individual locusts’ response to conspecifics and to other external stimuli (such as light, temperature and obstacles) in order to explain the generation and movement of hopper bands. We will also quantify how such responses are modulated by internal states variables, such as hunger and phase. We will use novel digital imaging techniques to automatically track the movement of locusts within relatively small groups in the laboratory (see image to the left, bottom), as well as within subsections of large groups in the field. The results from these experiments will be used to parameterise new mathematical models of locust movement. Mathematical models are a particularly important research tool in the study of collective animal behaviour because individual interactions are not reliably scaled up to population-level properties by verbal arguments alone. Indeed, while the data collected through digital imaging will be used to parameterise individual-based computer models of swarm formation and movement, another aim of this project will be to relate these individual-based models to mathematical models of pattern formation, such as partial differential equations and those found in statistical mechanics. We will build our understanding by providing descriptions and explanations at a variety of temporal and spatial scales. Integrating our understanding at these scales will then allow us a more coherent view of the whole system.

Information transfer / social learning

In the natural world animals that exhibit coordinated behaviour patterns can gather and process data from many distinct physical locations simultaneously. This decentralisation has important implications for the acquisition and processing of information within populations. A common problem that group-living animals are faced with is that only one or a few individuals have access to information about the location of a food source, a suitable nesting site or a predator, whereas the majority of the group do not. Thus the survival of the individual is often strongly dependent on the presence of other group members. Recently, much interest has been focused on the influence such ‘informed’ individuals have, and it has been shown that initially naïve individuals can acquire relevant information by interacting with knowledgeable individuals. Such acquisition of information is termed “social learning”.

To understand better the processes, and consequences of, social learning in animal groups I am developing computational techniques to quantify 'social network structures' within real animal groups (typically fish schools). Fish, in particular the guppy, are a good study organism since it is known that individuals may benefit from reciprocal altruism. Consequently an understanding of the network structures will improve our understanding of the evolution of cooperation in groups of unrelated (or largely unrelated) individuals, as well as other important ecological properties such as the spread of disease. This approach is being combined with fundamental models of animal aggregation (see Fish Schooling, above) in which patterns of clumping (social network formation) emerge from individual behaviour, allowing us to develop new spatial models of information/disease transmission grounded in empirical research through collaboration with Dr. Jens Krause and his fish research group at the University of Leeds, UK.

Current papers on this topic include:

Effective leadership and decision making in animal groups on the move, Nature

Computer vision

The way in which we perceive phenomena has a profound impact on our understanding of the world. Although humans can easily derive and express information from a scene in symbolic qualitative statements, we are subject to bias and inaccuracies when quantifying visual information. Digital image processing and analysis techniques, collectively termed ‘computer vision’, provide an alternative means of interpreting images and extracting information from them.

The study of animal behaviour often requires accurate recording of the movement of individuals from which properties such as the movement patterns, interactions among individuals and between individuals and their environment must be determined. There are many limitations to manual collection of data of this type. Humans can seldom accurately quantify the successive spatial locations of organisms necessary for behavioural analysis. It is also practically impossible to perform continuous recording, thereby true frequencies, and the times at which behaviour patterns stopped and started are seldom available. When detailed information at the level of the individual organism is needed, it is often impossible to observe more than one individual at a time due to the restricted focal range of the human visual system. The rate of movement, or behavioural transitions, may also occur over temporal scales that are unsuitable for human observation. I have developed digital imaging techniques to automatically track and record the behaviour of many organisms simultaneously (see Locust Swarms above, and image from the tracking of ants in the laboratory, to the left, bottom). Such technology is likely to yield novel and important insights into individual behavioural patterns, and the use of computer vision may benefit many areas of behavioural research.

Automated trajectory reconstruction and analysis will allow a unique insight into the mechanisms of grouping behaviour, and will provide essential data for the construction (and testing) of realistic computer models. It is also likely to provide important information to researchers interested in developing artificial visual systems, and those interested in developing motion control algorithms for autonomous mobile robots.

Through my computer vision company 'Eciton Software', I develop novel imaging solutions for international clients (see image to the left, top).