A random and universal architectural simulation

In this chapter we will see how the process of fragmented modelling and integrative computer simulation, which had originally been developed in order to simulate coffee-plant growth, would lead to an architectural simulation of plants that could be described as universal thanks to its ability to simulate, by simple extension, the entirety of plant architectures observed in nature. In contrast, I will reveal the technical limitations of the more classic theoretical or biometrical formal models, in particular when it comes to grasping extremely composite objects such as plants. Theoretical models in fact pay little heed either to the complexity of the living essence itself (reductionism) or to the evolving and intertwined nature of the optimization function that is meant to follow plant genesis (a reduction to optimization principles that homogenize and dehistoricize the scenario, despite its complex interweaving of cellular differentiation and of growth). For its part, statistical biometry requires simple models for a precise usage, without recognizing that it over-constrains its language, whereas it could be more generous with regard to the data without always reducing them to averages, variances and so on. Computer simulation, on the contrary, enables such generosity. Although computer simulation, like biometry, has the advantage of not viewing the plant as a theoretical object, it also allows a sort of underlying theoria by constructing a sort of multi-dimensional scale drawing as opposed to the perspectives represented by the models. We will see nonetheless that this search for realism was not always understood or well received by modellers of living beings - to the extent that it risked disappearing and falling into oblivion in the early 1980s. The detailed criticisms of earlier plant architecture and growth models that de Reffye expressed at the start of his doctoral thesis cannot explain the driving force behind his achievement in his research work. Reffye concluded that "'random coffee plants' are then obtained that have the same behaviour as the plant observed in the field". From a botanical point of view, the main success of de Reffye's work on universal simulation lies in its ability to simulate the entirety of de Francis Halle's and Roelof A. A. Oldeman's various architectural models. An "architectural model" in the sense intended by Halle and Oldeman is therefore fully defined when a particular combination of morphological characteristics and t