Optimization and Control in Procedural Modeling

2019-08-14T17:34:36Z (GMT) by Vojtech Krs
Procedural modeling is a powerful technique used in computer graphics to create geometric models. Instead of manual geometry definition, models are generated implicitly from a set of rules and parameters. Procedural systems have found widespread use in generating content for games, film, and simulation of natural phenomena. Their strength comes from the ability to automatically generate large amount of varied geometry. One of their drawbacks is lack of control because a small change in input parameters often causes large changes in the generated model.
In this work we present three novel procedural systems, investigate different forms of control, namely simulation and optimization, and discuss them in terms of general procedural modeling workflow. First we show modeling of 3D objects with arbitrary topology via erosion and deposition simulation controlled by Smoothed Particle Hydrodynamics. Next, we present an algorithm for generating 3D curves using 2D sketches and contextual geometry. Finally, we propose a novel procedural system capable of generating arbitrary type of geometry with respect to user-defined constraints.
We show that these systems can be controlled via several means and identify common preconditions that facilitate control: maximizing interactivity and amount of structured information input, minimizing unexpected behaviour, and local control akin to traditional modeling.