Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
Artists and architects often need to handle multiple constraints during design of physical constructions. We define a performative constraint as any constraint on design that is tied to the performance of the model--either during fabrication, construction, daily use, or destruction. Even for small to medium scale models, there are functional criteria such as the ease of fabrication and the assembly process, or even the interplay of light with the material. Computational tools can greatly aid in this process, assisting with the lower-level performative constraints, while the designer handles the high-level artistic decisions. Additionally, using new fabrication methods, our tools can aid in lowering the difficulty of building complex constructions, making them accessible to hobbyists. In this thesis, we present three computational methods for designing with different approaches, each with a different material, fabrication method, and use case. The first method is a construction with intersecting planar pieces that can be laser cut or milled. These 3D forms are assembled by sliding pieces into each other along straight slits, and do not require other support such as glue or screws. We present a mathematical abstraction that formalizes the constraints between pieces as a graph, including fabrication and assembly constraints, and ensure global rigidity of the sculpture. We also propose an optimization algorithm to guide the user using automatic constraint satisfaction based on analysis of the constraint relation graph. We demonstrate our approach by creating several small- to medium-scale examples including functional furniture. The second method presents a solution to building a 3D sculpture out of existing building blocks that can be found in many homes. Starting from the voxelization of a 3D mesh we merge voxels to form larger bricks, and then analyze and repair structural problems based on a graph representation of the block connections. We then output layer-by-layer building instructions to allow a user to quickly and easily build the model. We also present extensions such as hollowing the models to use less bricks, limiting the number of bricks of each size, and including color constraints. We present both real and virtual brick constructions and associated timings, showing improvements over previous work. The final case presented tackles the inverse design problem of finding a surface to produce a target caustic on a receiver plane when light is refracted or reflected. This is an example where the performative constraint is the principal driver of the design. We introduce an optimal transport formulation to find a correspondence between the incoming light and the output target light distribution. We then show a 3D optimization that finds the surface that transports light based on the correspondence map. Our approach supports piecewise smooth surfaces that are as smooth as possible but allow for creases, to greatly reduce the amount of artifacts while allowing light to be completely diverted producing completely black regions. We show how this leads to a very large space of high-contrast, high-resolution caustic images, including point and line singularities of infinite light density as well as photo-realistic images. Our approach leads to surfaces that can be milled using standard CNC milling. We demonstrate the approach showing both simulated and fabricated examples.
Rémi Guillaume Petitpierre, Paul Robert Guhennec, Beatrice Vaienti, Didier Louis Dupertuis