The University of Sussex

Methodological issues within a framework to support a class of artificial life worlds in cellular automata

Pedro Paulo Balbi de Oliveira

A brief description is made of a family of two-dimensional, non-deterministic cellular automata, whose temporal evolution in periodic background can be easily described in terms of the metaphor of an artificial-life world where worm-like organisms randomly roam around, sexually reproducing, interacting with the environment, and being selected. This family of automata, together with a methodology stating how to use it, and suggesting the most appropriate kinds of questions that it can be used to address, can be seen as a framework to define and support a particular class of artificial-life worlds. This class is characterised by organisms of arbitrary length, formed by sequences of contiguous cells, so that the cells at each end of an organism can be intuitively thought of as its head and tail, whereas the cells in between constitute its body. The dynamics of the automata rely upon four categories of states which represent the environment, the organism's terminal and body cells, and an additional category that allows the organisms to move. The emphasis of the presentation is on the methodology underlying the framework, rather than on the detailed description of the cellular automata themselves. It is shown that the dynamics of the automata can be conceived of as a "programmable" virtual machine. The idea is that, since the dynamics of the automata is entirely defined by the four state categories above, a "program" that is implemented on that machine, refers in fact, to the specification of a set of state transitions involving specific states in these categories. In order to illustrate the use of the framework, an in-depth example is given, namely, the simulation of a Turing machine through the interaction organism/environment. The possibilities and limitations of the framework are discussed.

This paper is not available online