Visual Instance-Based Modeling of Procedure Specifications for Architectural Structures

Under the pressure of competition and the expectations of consumers, game developers are in a race to create the biggest and most detailed worlds in their productions. Nowadays, thanks to the evolution of computer hardware, great amounts of detail to be put in those worlds, and this trend is not expected to end any time soon. But that detail has to be created by hand most of the time and that is becoming a very costly activity, to such an extent even that it is becoming unprofitable. Cannibal Game Studios develops software to make content creation easier and more cost-efficient, and from that company came the assignment for this work. This work focuses on the creation of one particular type of content: architectural structures. The goal of this work is stated as follows: Create an approach and a prototype system that will speed up the creation of architectural structures. The first part of the work consisted of researching two related subjects and developing an approach based on the findings of the research. The first is the domain of architectural structure creation. The research explores the process of architectural structure creation and the problems that exist in it. The development of a new approach should not only solve the problems, but it should also fit in with the existing processes and design principles. Through interviews with game developers in the Netherlands, an understanding was gained of these matters and a set of requirements was created that the approach should fulfill. The second subject of research pertains to the existing approaches that have been developed to solve the identified problems. The most relevant existing approaches are based on procedures and a selection of those approaches is reviewed in order to discover their strengths and weaknesses. While all approaches have their benefits, they also exhibit several shortcomings. With the development of a new approach that satisfies the requirements of the games industry, it is important to use their benefits in our advantage and to keep their shortcomings in mind in order to provide a more complete solution. This led to the inception of an instance-based modeling approach that allows users to specify procedures for the design of a structure through (real-time) interaction with a special instance of that design. In this approach, designs are specified and instances of the design can be created by a procedural system. The approach defines structures in terms of components and relations between those components. The instantiation process for the instance of a structure instantiates, forms, and positions instances of components based on their relation definitions to other components. The approach uses patterns to govern the repeated instantiation of components. Patterns are used in a similar fashion to select component instances as parents for instantiated connections relations. One of the most interesting concepts of the approach is that it allows components to have non-hierarchic relations to several other components. This approach shields the user from coding abstract rules and presents a familiar 3D modeling interface instead. In the second part of this work, a prototype system was implemented based on this new approach. This was done based on the software framework of Cannibal Game Studios, which includes a 3D game Engine and a content editor framework. For the prototype, it was decided to use connections as the relations between components of a structure, and curves as the component shape type to use. During the development of the prototype, it became apparent that the procedural instantiation logic was much more complex to create than anticipated. It proved to be the biggest challenge in the project to keep instances of a component up-to-date in real-time when the design of the component was modified. The result of this work is in the first place a new approach for Computer-Aided Design for architectural structures that allows users to create procedural structure designs in a non-technical way. It also allows users to easily create and work with non-hierarchical relations between components in the design. In the second place it has led to the creation of a prototype system that serves as a proof of concept for this approach. Although the prototype is not complete enough to be considered a usable product, it does prove that an implementation of the proposed approach can actually be realized. Aside from the complete approach, even the individual concepts of the approach that were implemented in the system have value to the content creation process. Future work can be performed in two ways. The first is to improve the current implementation by making it more efficient and by adding more features. These features would include more interaction functionality to make the system more user-friendly, and more functional features to make it more flexible and complete. Some examples of the latter are the implementation of more shape types and more relation types. The second way in which future work can be done is to develop the approach itself further. For instance, other ways of relation specification can be explored on both the technical and the user interface level.Computer Graphics and CAD/CAM GroupMediamaticsElectrical Engineering, Mathematics and Computer Scienc