Generative Design in Civil Engineering Using Cellular Automata

Rafal Kicinger

George Mason University

In this presentation, the concept of generative design and its applications in structural and transportation engineering are discussed. The idea of generative design is based on the use of automated mechanisms for producing design concepts of often complex engineering systems. Several types of such mechanisms for generative design have been proposed and studied, including rule-based systems, shape grammars, and cellular automata. In this research project, one- and two-dimensional cellular automata were used to generate designs of steel structural systems in tall buildings and traffic control systems in urban areas. These generative mechanisms were encoded in so-called generative representations that were manipulated by search and optimization algorithms in order to discover interesting and optimal designs in these large design spaces.

Generative representations based on one- and two-dimensional cellular automata were implemented in a computer design support tool. This tool was subsequently used to conduct a large number of computational experiments in which interesting and optimal designs were sought. The optimality of generated designs was determined by the total weight (a good estimate of the cost) of steel structural systems and the total vehicle travel time in the case of traffic control systems.  On the other hand, visual characteristics of generated patterns were used to determine qualitative differences among generated designs and traditionally known patterns from civil engineering literature.  

Results of extensive computational experiments have shown that cellular automata based generative representations produce interesting and novel structural patterns. When combined with evolutionary algorithms, they also produce optimal solutions significantly faster than traditionally used parameterized representations.

[presentation materials]