Logo[ Bristol CS | Index | MLBC Group | Tim Kovacs | SISEBIA ]

Social Insects, Simulated Evolution and Biologically Inspired Algorithms

Funded under the joint EPSRC/BBSRC call "Adaptive and Interactive Behaviour of Animal and Computational Systems", this joint research grant at the University of Bristol's University Research Centre in Behavioural Biology, involving the School of Biological Sciences and the Department of Computer Science, aims to study the evolution of individual and collective behaviours in social insects, and possible applications of those behaviours in computer science. The grant commenced on January 5th 2004 and ran for two years.

Abstract

Social insects employ ingenious and sophisticated algorithms to assess a variety of attributes of potential new nest sites, and to achieve a collective decision and coordinated emigration. Using methods from artificial life, we will study these house-hunting behaviours and their evolution in ants and honeybees. This research lies at the interface between computer science and biology, and provides exciting benefits to both disciplines. For biology, simulating the evolution of behaviours and their alternatives enables us to assess the adaptive benefit of one behaviour over another and thus to understand why a particular behaviour has evolved. This is not possible by simply analysing extant behaviours; simulation studies are crucial to allow comparisons with the behaviours as they could have been or indeed were in the ancestors of existing species. For computer science, social insects are a source of fascinating and unique insights for distributed decision-making systems in general. Social insects can inspire new solutions to many current problems in computer science for which robust solutions are difficult or impossible to find using traditional approaches, for example search problems in machine learning, control problems in communication networks and scheduling problems in distributed computing. We will develop a novel biologically inspired algorithm based on nest assessment and emigration behaviours, the characteristics of which will be decentralised decision making with an adjustable speed/accuracy trade-off. The design of this algorithm will be improved by the deeper understanding of these behaviours gained through the evolutionary behavioural modelling component of the research.

Grant Participants:

Publications

Media

Models


Tim Kovacs, Last modified on June 23 2013