Rule-based systems and in particular L-systems provide possibilities for modeling plants in extremely compact ways. The fundamental formalism of L-sys – tems is simple and has its aesthetic attraction. using various extensions, such as context-sensitive, parameterized, differential as well as open L-systems, Prusinkiewicz et al. succeeded in adding to formalism a wide range of options that allow for numerous branching patterns to be modeled. In [197] it was shown that with L-systems principally all 23 tree architectures of Halle, Oldeman and Tomlinson [81] can be simulated. Together with the additional functions that can be entered by the user, and hence the option of interactive modeling, the modeling process became a much easier task.
Also the visual results converged over the course of the time with the quality of procedural methods. Today models of similar quality can be produced with either approach. It should also be emphasized that the extraordinary spectrum of possible plant models is not restricted to trees. During the development of the discussed methodology, all types of plants were modeled. This differentiates the approach clearly from the procedural methods that were designed especially for particular structures.
However, the modeling aspect is still problematic. As already previously noted in this chapter, the production of a given plant with an L-system is a difficult process. The user must be proficient in order to represent a plant from the very start using an L-system. Many aspects of the plant have to be defined locally, and already small changes sometimes cause a complete modification of the total shape. In addition, the production of plant geometry is a computationally expensive task. After each parameter change, a complete expansion process must be worked through, including the geometry production, which significantly slows down progress. This is especially disturbing when working with huge models such as trees. However, with the improvement of computer efficiency, this becomes a less significant factor.
Likewise, the graph-based procedures discussed above do not contain any improvements with regard to the modeling options of rule-based procedures. With these methods, only part of what is definable with the L-systems is actually converted. The possibility of representing the topology of plants over graph structures is, however, an interesting aspect. The “rules” in this case denote instancing dependencies in the graph: if a node is arrived at, and if appropriate geometry is created, then transformations of the local coordinate system are invoked, and the geometries of all child components connected with the node are produced.
In the next chapter this mechanism is extended in order to store in a graph structure a part of the plant topology. The nodes of the graph are now com-
ponents that contain data and algorithms in order to procedurally create parts of the plants. The combination of a graph-based rule system and procedural methods combines the advantages of both methods.
