Speaker
Description
With the rapid development of artificial intelligence technologies, intelligent design has become a popular research approach in recent years, enhancing design diversity, flexibility, and efficiency through intelligent algorithms. The self-organized design of block forms is a key component of intelligent urban design, aiming to optimize urban spatial layouts through adaptive mechanisms and algorithmic optimization. Based on the theory of spatial self-organization, this paper proposes a reinforcement learning-based method for the self-organized generation of block forms, enabling the system to learn optimal strategies through interaction with the environment.
This study consists of three parts. First, using Nanjing, China, as a case study, geometric calculations are applied to extract the spatial topological relationships of block forms, which are stored in a design prototype database. Building upon this, a reinforcement learning algorithm is introduced with a reward mechanism during the design process, considering various design constraints in urban space, such as development intensity, building density, building height, and green space ratio, to establish a self-organized design framework for block forms. Finally, three representative urban blocks in Nanjing are selected for generative design experiments. Based on real-time feedback from the reward function, dynamic adjustments and optimizations are made to the block forms, achieving a comprehensive optimization of economic, ecological, and social benefits while meeting design constraints. Unlike traditional rule-driven design methods, the approach used in this study involves the iterative testing and optimization process of the reinforcement learning model, allowing block form generation to not only adapt to the current environment but also evolve over time to address potential future demand changes. The effectiveness of this method is validated through a specific case study in Nanjing, providing new technical support for future intelligent urban design.
| Keywords | Block form; self-organized generation; reinforcement learning; intelligent optimization; urban design |
|---|---|
| Best Congress Paper Award | Yes |
