A developmental system for organic form synthesis
thesisposted on 31.01.2017, 04:27 by Porter, Benjamin
In many applications of computer graphics it is necessary to model natural forms. Worlds populated with both familiar and exotic flora and fauna, for example, are commonplace in animated films and computer games. Modelling organic forms using traditional computer-aided design or animation tools, however, is often a tedious and time-consuming process, particularly when constructing forms that grow and develop large amounts of complexity from simple beginnings. Procedural modelling, an alternative to the traditional, or "manual", approach to geometric modelling, seeks to address the problem of model complexity. Under this approach, a user specifies a procedure -- often with associated parameters and initial conditions -- which the computer then executes in order to construct a model. Procedural modelling systems are capable of generating extremely complex structures, such as sprawling landscapes, cityscapes, and forests. If biological or organic shapes are to be modelled, it is natural to consider the processes that occur within biological development. Developmental systems, based on aspects of biological development, have emerged as a powerful technique for generating a rich variety of organic forms; however, there remains a number of forms that existing systems are unable to generate effectively. This thesis introduces the Simplicial Developmental System (SDS), a system capable of automatically generating developing organic forms that are difficult, or impossible, to create using existing methods. These forms can be characterised in natural language as "organic, smooth, soft, squishy, and modular". SDS integrates a number of elements, such as cellular behaviour, soft-body mechanics and morphogen diffusion. These elements operate on a unified, adaptive, geometric representation of volumetric form -- the simplicial complex -- that is embedded within a physically simulated environment. The thesis presents two instances of SDS: a system for generating 2D forms with a triangular mesh representation, and a system for generating 3D forms with a tetrahedral mesh representation. Modelling biological development on these representations raises a number of questions; for instance, how does the division of a cell modify the tetrahedral mesh? This research offers a solution to this and other technical challenges. The capabilities of SDS are exhibited in a number of experimental results presented in the thesis. These experiments demonstrate how SDS can be used to generate continuous sequences of developing organic forms. Two biological models of growth were successfully implemented in the 2D system: limb bud development in chicken embryogenesis and antero-posterior segmentation in Drosophila melanogaster. Experiments with these models demonstrate how cells in SDS can coordinate their behaviour in order to generate higher-order structures and patterns. This thesis also presents a number of experiments performed with the 3D system, demonstrating the effectiveness of SDS in automating the generation of complex 3D geometric models with organic features such as smooth surfaces, modularity, environment-sensitivity, similarity with variation, and elastic deformation.
Awards: Winner of the Mollie Holman Doctoral Medal for Excellence, Faculty of Information Technology, 2011.