Evolution in Asynchronous Cellular Automata

摘要

Building on the work of Von Neumann, Langton, and Sayama among others, we introduce the first examples of evolution in populations of self-reproducing configurations in asynchronous cellular automata. Reliance on a global synchronous update signal has been a limitation of all solutions since the problem of achieving self-production in cellular automata was first attacked by Von Neumann half a century ago. Results of the author obviate the need for this restriction. We review our simple constructive mechanism to transform any cellular automata network with synchronous update into one with essentially the same behavior but whose cells may be updated randomly and asyn-chronously. The generality of this mechanism is guaranteed by a general mathematical theorem that any synchronous cellular automata configuration and rule can be realized asynchronously in such a way that the behavior of the original synchronous cellular automata can be completely recovered from that of the corresponding asynchronous cellular automaton, in which temporal synchronization locally stays within small tolerances. It follows that most results on self-reproduction, universal computation and construction, and evolution in populations of self-reproducing configurations in cellular automata that have been obtained in the past carry over to the asynchronous domain using the method described here. Here we discuss requirements for evolutionary systems in cellular automata and describe implemented examples of our procedure applied to a variety of self-reproducing systems (Byl, Reggia et al., Langton, Sayama). In particular, we have implemented Sayama's evo-loop system asynchronously, giving the first example of evolution in asynchronous cellular automata.