Optimal Design of Urban Public Space Landscapes Integrating AI Technology

Urban public spaces reflect urban vitality and governance quality, but traditional landscape design lacks real-time adaptability to diverse needs. This study presents a “perception–reasoning–generation–feedback” framework integrating computer vision, reinforcement learning, and multi-objective evolutionary algorithms. It models crowd dynamics, microclimate, and emotional semantics, validated in a waterfront plaza. Multi-source sensors collect spatiotemporal data; Vision-Transformer-RL identifies behaviors and predicts demand. Adaptive operators generate layout options, refined interactively. Results show enhanced functionality, satisfaction, and energy efficiency, highlighting AI's potential in landscape design. The approach supports the shift from static plans to dynamic algorithms and offers a human-centered perspective for smart cities. Future work may integrate AR co-creation and carbon assessment, promoting sustainable, inclusive urban development.