Sci Rep. 2025 Nov 21;15(1):41255. doi: 10.1038/s41598-025-25143-9.
ABSTRACT
Traditional static design approaches struggle to address the dynamic environmental conditions and evolving user needs of contemporary urban open spaces. This research proposes a comprehensive AI-driven real-time responsive design methodology that integrates multi-modal sensing data to enable dynamic optimization of urban open spaces. The proposed framework employs a hierarchical data fusion architecture that processes heterogeneous sensor streams including visual, acoustic, and environmental data through advanced machine learning algorithms. Deep learning-based spatial optimization models combined with reinforcement learning mechanisms generate adaptive design solutions that respond to real-time conditions while maintaining design quality standards. The system achieves sub-100ms response times through optimized computational architectures and intelligent caching strategies. Experimental validation conducted across three representative urban sites demonstrates significant improvements including 34.2% increase in space utilization efficiency (measured as the ratio of actively used area to total available space), 28.7% enhancement in pedestrian flow optimization (quantified through movement speed and path directness metrics), and 22.3% reduction in operational costs compared to conventional static design approaches. The practical application case study at Metropolitan Central Plaza, a 2.4-hectare transit-oriented public space in Shanghai’s dense urban district, validates the methodology’s effectiveness in real-world deployment, showing substantial improvements in user satisfaction metrics and environmental quality indicators. This research establishes foundational principles for developing intelligent urban environments that can continuously adapt to changing conditions while optimizing resource utilization and enhancing user experience quality.
PMID:41271992 | DOI:10.1038/s41598-025-25143-9
