Computer-Aided Architectural Plan Generation: 2D Mesh Segmentation, Search Optimisation, and Clustering for Generating Plans

In the architectural design, segmenting given floorplates into spatial partitions and subsequently deriving arrangements of architectural elements (such as walls, furniture, and entries, etc.) to form the so-called architectural plans is a complex problem. Each spatial configuration involves various constraints and variables (such as unit types, areas, and orientation) that need to be balanced. Architectural plans represent the core ideas and solutions in architectural communication. However, in practice, resolving architectural plan solutions and managing the solutions into one plan takes a significant amount of effort during the phase of schematic design. This is because the solution of spatial allocation cannot easily be procedurally derived, and the problems and constraints of the task are often ill-defined. Typically, the solution evolves over several iterations. This process requires larger numbers of solutions to be made. To address these issues, a systematic method is demanded.

This thesis interrogates a computer-aided architectural plan generation method using a parallel graph-search algorithm in combination with an evolutionary optimisation algorithm. This research tests architectural plan generations in residential design, particularly targeting the typology of the residential tower in the case study.