Computational Design of Residential Units’ Floor Layout: A Heuristic Algorithm

Shurui Yan,Nianxiong Liu
DOI: https://doi.org/10.1016/j.jobe.2024.110546
IF: 7.144
2024-01-01
Journal of Building Engineering
Abstract:Computer-aided intelligent design can improve the quality and efficiency of residential building design. However, the design conditions are diverse, and numerous combinations of spatial elements exist. We propose a heuristic algorithm-the combined Monte Carlo tree search (MCTS) and particle swarm optimization (PSO) algorithm (PSO-MCTS)-for the automated design of residential floor layouts. The MCTS algorithm deals with discrete variables (room locations). It considers architectural design experience and design standards during node pruning to compress the search space and simplify the computation. The PSO algorithm deals with continuous variables (room sizes) and accelerates the computing speed due to parallel processing. The PSO-MCTS algorithm achieves a good balance between computational efficiency and accuracy. It is suitable for various design conditions and requirements. The input parameters include the daylight fa & ccedil;ade, boundary scope, entrance location, room type and size, and relationships with adjacent rooms. The PSO-MCTS provides 100 %, 80 %, and 75 % better results than heuristic algorithms, designs by architects, and a state-of-the-art deep learning model, respectively.
What problem does this paper attempt to address?