AI algorithm-driven generative technology is currently a cutting-edge field in environmental design. The study analyzes the principles and characteristics of neural networks and generative adversarial networks from the perspective of the principles of deep learning technology, and explores their applications in the field of environmental design. Subsequently, based on the generative adversarial network model, we design an environment design generation model with improved generative adversarial network. With the mechanism of moving window, the attention module in the model is improved to construct a hierarchical generative adversarial network model. Finally, the application and impact analysis of the AI generative model in environmental design are explored in the experiment. Through the case study of a historical neighborhood, in the comprehensive analysis of spatial aesthetics, the scores of “positive” scenarios are higher than those of “negative” scenarios. Meanwhile, the overall coordination is an important factor affecting the visual quality of spatial aesthetics.
