.png){kind=icon}
.png){kind=icon}
Return
A knowledge-informed multi-agent reinforcement learning approach for cost-effective aesthetic facades generative design
.png){kind=icon}
PRE
AI
B
W
Z
H
DOI：10.1016/j.aei.2026.104743.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Building façades critically influence aesthetic appeal and construction costs, yet achieving cost-effective aesthetics in façade design remains challenging due to subjective evaluation and iterative work to balance conflicting multi-stakeholder requirements. This paper proposes a knowledge-informed multi-agent reinforcement learning (K-MARL) framework to enhance cost-effective aesthetics in façade design. First, a parametric model of building facades is set to define key design variables. Three knowledge-informed surrogate models are then trained for quantitative predictions of façade aesthetics from architectural, residential, and urban planners’ perspectives. Third, a decentralized cooperative MARL integrated with pre-trained surrogate models is developed for simulating design negotiations via weighted action aggregation and utility-based consensus mechanisms. Finally, multi-agent negotiations are conducted across various cost-biases, producing Pareto optimal solutions that represent performance trade-offs. The method was evaluated in a real-world case study in China. Results show that the optimized solutions can improve up to 9.61% in aesthetics score and save up to 3.8% in construction cost compared to the baseline plan. The study advances agentic generative intelligence in façade design by demonstrating a deployable framework for performance-driven design solutions, and provides interpretable, actionable design insights for cost-aesthetics balanced façades.
Keywords:
facade design
multi-agent reinforcement learning
cost-effective aesthetics
parametric modeling
Pareto optimal solutions
Journal
IF:
9.9
Papers:
4.0K
Citations:
1.7W
Organization
No organization information available
Cited Papers