The International Association for Automation and Robotics in Construction
Publications / CCC 2025 - Zadar, Croatia
GENERATIVE DESIGN FOR SITE LAYOUT OPTIMIZATION
Construction site layout planning (CSLP) is a strategy and decision-making process for the placement of temporary facilities and areas within a construction site. CSLP initiates and affects the direction and productivity of projects. CSLP often encounters several challenges brought about by uncertainties and unquantified decisions. Research efforts to automate CSLP have evidently faced insufficient field validation and adaptation. In literature there has also been difficulty in considering three-dimensional spaces and movement for site layouts. The study aims to contribute to the automation and applicability of quantified three-dimensional layouts in CSLP. Generative Design is a technological tool that intakes numerical inputs and produces computational and model outputs in real time. A realistic consideration of construction projects requires dynamic layouts that consider three dimensional spaces. Grasshopper Generative Design software was utilized to optimize layouts minimizing an objective function in a quadratic assignment problem. It optimizes using the Genetic Algorithm, and its effectiveness was demonstrated by testing against benchmark results. A site layout optimization case was conducted on a 12-storey educational building with three phases. Furthermore, a qualitative interview with the site planner was conducted to establish a contrast between industry practice and theory. The use of benchmark data in this optimization process produced a decreased minimum cost by 17% compared to the original publication. Results produced rapid generation of three-dimensional models and layouts for exploration and visualization. The case study demonstrated the applicability of the technique by producing an optimized three-dimensional layout for distinct phases of the project. Aside from this, the outputs were found to be readily interpretable by an industry professional due to its visuals and objective driven layouts. Future works may use Generative Design plug-ins for multi-objective optimization and physics simulations such as obstacle detection.