Classifier Systems, Cellular Automata and Computational Intelligence: Applications in Multidisciplinary Synthesis

Abstract

This paper describes new developments in numerical methods for multidisciplinary analysis and design that belong to the general category of computational intelligence techniques. The basis of these methods is in the use of information obtained from the computational domain to enhance the efficiency of multidisciplinary design procedures. The paper briefly documents the adaptation of artificial neural network based procedures for function approximation in the design process. The larger focus is in the discussion of two newer methods, classifier systems and cellular automata, and their applications in multidisciplinary analysis and design. Classifier systems are shown to be an evolved form of the classical rule based systems, where information from the computational domain is used to derive new rules for subsequent operations, Cellular automata approach is presented as an efficient method for analysis and design for massively parallel computers. The focus here is on the use of evolutionary computing methods to discover rules for automata evolution. Numerical results are included in support of these concepts.