Applications of Evolutionary Computing to Behavioural Ecology Extended Abstract

This work concerns a study of the use of adaptive computational approaches as a mechanism for individual agent based modeling of adaptation in ecological systems, particularly in situations which are unsuited to conventional analytical techniques (ie. using continuous mathematics). In this context we define the term Computational Ecology to mean the software simulation or model (derived from some observed biological system) comprising an environment and inhabited by autonomous agents. Behavioural ecology is generally concerned with the behaviour of animals both individually and in groups. Population biology, on the other hand, concerns itself with with the effects on the population as a whole of environmental and biological changes. Traditionally, neither address the demographic impact of behavioural changes, or vice versa. Recently there has been a move towards investigating the effects of individual behaviour on population dynamics (Hassel and May 1985), much of this work employs analytical mathematical methods. It would appear intuitively evident that there are ecological circumstances under which the consequences to an individual of its own behaviour may be dependent on the behaviour of the other individuals that also inhabit its environment, this may be as a result of, for example, individualindividual interactions or the effect of the population members’ behaviour on the environment itself. An example of the convergence of behavioural and population ecology is the consideration of the relationship between Evolutionary Stable Strategies (ESSs) (Maynard-Smith 1982) and population dynamics (Parker 1985). Game theory and ESSs in particular are prime candidates for large scale computer modelling on the basis of the explicit representation of individuals. It is possible that adaptive computing agents coexisting in an environment may display the emergence of ESSs. The success of some individual based fieldwork (eg. Huston et. al. 1984) provided ample motivation for individual based simulation. However, it was clear that any system implemented would necessarily embody a very high degree of abstraction, yet to be useful should still preserve the essential relationships. The intention was to implement explicitly the individual members of the population as a complementary approach to the more conventional techniques of population biology using adaptive computational techniques and observe the evolution of strategies, if any, over time, given the broad brush nature of such an implementation it was considered best to look for strategic extremes and in this context K and r selection effects appeared to be the most likely targets. K refers to the carrying capacity of the environment and r the rate of increase of the population. Elephants provide a good example of K-strategists which live in stable habitats and adapt towards maintaining population at an equilibrium and improved intraspecies competitive ability, they are characterised by larger size, long lifespan and low r. In contrast r-strategists colonise habitats of a temporary nature and their population variation follows ‘boom and bust’ cycles, high fecundity (i.e. high r), small size and short life spans characterise the