Study of Structural Optimization Technique Using Evolutionaly Cellular Automata.

Abstract

Cellular Automata(CA) are examples of dynamical systems which exhibitself organizingbehavior with increasing time. They are useful in modeling modular systems. One such application of modularity is described in this paper where a structural plate is considered as composed of smallerstructural moduleswhich are considered as cells in a lattice of sites in a CA and have discrete values updated in discrete time steps according to local rules. These local rules are generally fixed in a CA, but we consider these rules as evolvable. To evolve the local rules, we use the Genetic Algorithm(GA) model. In this paper, two structural optimization methods by Evolutionary Cellular Automata(ECA) are presented. ECA is a Cellular Automata(CA) which is optimized for computation of certain problems by the Genetic Algorithm(GA). The first method is calledDirect Rule Encodingand the other is calledIndirect Rule Encoding. Direct Rule uses configuration of the neighborhood state as the CA's input. Indirect Rule uses theRatiobetween given cells' stress and sum of neighborhood stresses. These two methods are applied to minimize weight design problems and adaptive solutions are obtained From this experiment, we observe that Indirect Rule is effective in large scale problems.