- Represent an apartment blueprint by a polygon mesh using the half-edge data structure
- Generate smart mesh grids. A meaningful grid of the apartment should be able to represent as much viable partition of the apartment as possible.
- Generate walls and space partition of the apartment, according to grid, user specifications, and architectural quality via a genetic algorithm.
- we use the typing library to enable type checking and pycharm magic
- we enforce pep8
- we aim for a pylint note > 9
- we use 100 characters column width (not 120 nor 80)
We aim to use as few external libraries as possible. The project currently requires only Numpy, Shapely and Matplotlib. We use Pytest for testing and Pylint for linting.
-
correct snapping and cutting methods to prevent the creation of half-edge with the same starting and ending vertex -
change the way the enclosed face are connected (used orthogonal projection on closest edge) -
check failing floor plan : Noisy_A145, Antony_B14 (hard fails), Bussy_Regis, Massy_C102 (soft fails) -
add load bearing walls import (note some plan fail due to improper geometry : duct overlapping load bearing wall) -
create grid generator - add way more unit tests (work in progress)
- add matplotlib live visualization with pytest debugging
- create a parallel cut (slice), useful for non rectilinear apartment
- add space generation from seed growth (the idea is to generate space by growing them from a seed point)
- create space mutations : one face, whole border, whole perimeter etc.~
- create space cost functions
- create a crossover function enabling to mix two blueprints for the same apartment
- create a repair function : that will insure correct apartment circulation
- create genetic algorithm framework : population generation, random crossovers, random mutations, repair population, cost function evaluation, selection of new generation
- replace sanity checks that raise ValueError exception with assertions