Using evolutionary algorithms for designing novel 3D objects

There are numerous varieties of living designs specialized for every habitat in nature. Resolving these designs' maturation mechanisms starting from the replication of one initial cell (zygote) can help us design objects that we need. In order to understand how a big organism emerges from millions of tiny cells, cell behaviors must be understood sufficiently. Cells read a given instruction booklet -the DNA- according to the situation that they are in, and they decide what to do with the rules coded into the booklet. If these rules, how they are encoded in the DNA and which mechanism is employed to decode them are known, the same methods can be applied to the computerized 3D design. In this thesis, how multicellular creature embryos are maturated in the womb or egg is investigated, thereafter performance of this mechanism on producing 3D artificial novel objects is analyzed via making a simple simulation of that. An application developed with Unity3D game engine starts with a set of symbolic cubic zygotes and DNAs. In order for the artificial embryos to reach the ultimate maturity, the cells question their DNAs to reproduce, suicide, protein synthesis etc. to carry out their actions. Embryos having random genes are evolved to form optimal designs for given tasks by the help of a "steady state genetic algorithm". In order to evaluate the ability to create original 3D objects of this system known as "artificial embryogeny", a number of tests are performed. As a result, it is observed that proposed evolutionary algorithm is able to generate shapes that resemble given target shapes which have different characteristics, with a similarity between 50% and 85%. The system proposed in this study can easily be adapted to many different design problems simply by adjusting the fitness function.