AI-driven energy-efficient floorplan generation is a transformative tool addressing energy challenges in building design driven by climate change, carbon emissions, and rising energy demands. This review explores Automated Floorplan Generation with Energy Efficiency Optimization (AFG-EEO) methodologies, integrating design generation, performance evaluation, and optimization to streamline architectural processes. It highlights gaps in methods, emphasizing user-centric enhancements, architect involvement, and future directions to improve methodologies, offering researchers, practitioners, and stakeholders valuable foresight.
Over the last decade, climate change has accelerated, causing a faster temperature increase than in the previous century. People spend more than 90% of their time indoors. There is a strong relationship between indoor environmental quality demand and carbon emission. On the other hand, the building sector consumes 40% of global energy consumption, which is expected to rise due to the impact of climate change and population growth. Moreover, the building sector is particularly concerned, responsible for a substantial 28% of total energy-related CO2 emissions, mainly from electricity generation for building use. However, the design and construction of high-performance buildings have the potential to reduce energy consumption and Greenhouse gas emissions as reported by the International Energy Agency by 30% by 2050.
For this international trend, architectural design decision-makers should pay more attention to using energy simulation, artificial intelligence (AI), and automated floorplan generation (AFG) tools to achieve high-performance buildings and net-zero energy consumption. Integrating AI into design processes represents a paradigm shift in the architectural field. This shift marks a fundamental change in architectural design's underlying principles, methods, and practices. Traditional sustainable design process typically begins with a detailed climate analysis, followed by multiple sketches to incorporate these findings, which is a time-consuming sequence. Subsequent steps involve manual iterations of floorplans and sections, tested through simulation tools, and repeated until a satisfactory design is achieved.
In contrast, AI tools have conducted a revolutionary approach that can rapidly generate comprehensive sets of floorplans that adeptly address climate concerns and functional configurations and yield high-quality renders within minutes with a mere click or even a textual input devoid of specific climate analysis, sketches, or predefined design requirements. In the meantime, it is essential to note that the integration of AI does not aim to replace architects with machine learning models; instead, it empowers them with a highly advanced foundation, eliminating the need to reinvent the wheel for each design. This paradigm shift represents a significant deviation from the traditional design process, offering architects a progressive starting point for their creative endeavors.
A lot of AI and AFG tools exist at architects' disposal. However, a critical observation shows that most of these tools are available in the market and prioritize factors such as maximizing floor space and facilitating fast design generation, typically catering to the real estate sector's needs.
One particularly promising application is using AI algorithm tools to generate energy-efficient floorplans. These tools still need to pay more attention to environmental considerations such as energy consumption, health, and well-being due to a lack of necessary parameters such as sun path diagrams, exterior temperature range, wind direction, and speeds. The floorplans may be spatially efficient but need environmental consciousness.
Several research prototypes have explored AI tools as promising candidates for architectural design despite their limited availability in the market (see Table 1). These prototypes emphasize generating functional floorplans and building masses that comprehensively account for factors such as energy consumption, daylight, Universal Thermal Comfort Index (UTCI), and carbon emissions, reflecting a comprehensive approach to architectural design that seeks to balance functionality and environmental responsibility. Currently, the market offers a variety of energy performance simulation tools, including Covetool, OpenStudio, Honeybee, BEopt, eQuest, and EnergyPlus. However, it is important to note that these tools primarily serve as evaluation instruments rather than generative tools. They require pre-existing architectural models for performance assessment instead of independently generating energy-efficient floorplans or building designs.
Both research prototypes and market-available automated floorplan generation tools have distinct advantages and limitations. Market tools can handle large-scale projects, such as schools, healthcare facilities, and mixed-use developments. They can produce both 2D floorplans and 3D masses with detailed layouts. In contrast, research prototypes are typically confined to small residential floorplans and often focus on 2D outputs, with limited forays into 3D modeling, which remain less advanced compared to market tools.
A key strength of research prototypes is their emphasis on environmental considerations, such as energy efficiency, carbon emissions, and thermal comfort, that are integrated throughout the design process. Conversely, market tools treat environmental factors as supplementary, often limiting their analyses to basic parameters like sunlight hours, radiation, and daylight. Such analyses may lack depth due to their reliance on less developed engines compared to tools like EnergyPlus and OpenStudio. Moreover, market tools generally introduce environmental assessments at the final evaluation stage rather than embedding them in the generative design process.
However, research prototypes face notable constraints, including limited hardware and software capabilities compared to the robust functionality of market tools. Market tools are designed with user-friendly interfaces and workflow integration in mind, supported by extensive documentation and machine learning models, making them accessible to a broad user base. They are also optimized for speed, enabling the rapid generation of multiple design iterations and seamless integration into fast-paced workflows. In contrast, research prototypes are more complex and require significant technical expertise to run. Also, they are slower due to computational requirements and limited performance optimization.
Another advantage of market tools is their seamless integration with established industry software such as Revit, Rhino, and AutoCAD, facilitating efficient data exchange and visualization. Research prototypes, by comparison, are frequently restricted to specific platforms, such as Rhino and Grasshopper, which can limit their adaptability and compatibility with broader industry workflows.
The goal of integrating AI applications for the automated generation of energy-efficient floorplans is not to supplant the role of architects, which remains irreplaceable. Instead, the intent is to streamline the design process.
Traditionally, architects invest substantial time comprehending climate patterns and incorporating them into preliminary designs. Subsequently, they expend considerable effort using various tools to analyze energy consumption, daylight availability, carbon emissions, and material properties, iterating their designs based on each simulation's outcomes. This process is often repeated numerous times.
AI-driven automated floorplan generation serves as a time-saving catalyst by providing architects with energy-efficient preliminary schematic designs as a robust starting point. Architects can then optimize and refine these designs, applying specific simulation software such as special daylight autonomy (SDA) and annual sunlight exposure (ASE), also known as glare, with confidence, knowing that the generated floorplan represents the best-case scenario -- effectively streamlining what would otherwise be a time-consuming and iterative traditional process.
