Researchers from Tsinghua University in China have unveiled an AI-based urban planning system that showcased remarkable capabilities, outperforming human experts in crafting urban designs.
Chinese researchers have unveiled an AI-based urban planning system that showcased remarkable capabilities.hiLiuyun from Pixabay
AI Crafts Urban Plan
Traditionally, Science X Network reported that cities evolved organically, with immediate needs dictating development, often resulting in suboptimal outcomes.
The study noted that more recently, a shift towards a more systematic approach has emerged, focusing on creating designs that account for factors like livability and pollution control.
This evolution gave rise to urban planning as a distinct discipline, addressing the complexities that arise as urban developments increase in scale. In this study, the team harnessed AI to alleviate the complexities of urban planning.
Their artificial intelligence (AI) model was built around the "15-minute concept," aiming for residents to access essential services within a 15-minute radius. This approach, enhancing quality of life through reduced travel time and lower pollution levels, formed the system's foundation.
The AI was trained on existing human-generated plans and key design features, including green spaces, cycling paths, parks, and recreational areas. The system's first task was designing a compact community within a 3x3 grid city block.
This initial experiment enabled fine-tuning and subsequent improvements. As the scope of the development expanded, the system's proficiency persisted.
Results revealed that the AI-generated plans were on par, if not superior, to those crafted by human experts. Moreover, the time investment plummeted dramatically, from hours to mere seconds.
The team emphasizes that their system is not intended to replace human planners but to liberate them from routine tasks, enabling a more concentrated focus on overarching concepts.
The study introduces an AI-driven urban planning model that employs graph theory to depict diverse urban topologies. It formulates planning as a sequential decision-making challenge on this graph.
The team implemented a reinforcement learning model grounded in graph neural networks to address the vast solution space. Experiments encompassing both simulated and real-world communities demonstrated the computational model's superiority over plans devised by human experts based on objective metrics.
It exhibited the capacity to generate spatial plans tailored to various circumstances and requirements. The study also advocates for a collaborative workflow, integrating human expertise with AI capabilities to enhance productivity in developing efficient spatial plans.
"We also propose a human-artificial intelligence collaborative workflow of urban planning, in which human designers can substantially benefit from our model to be more productive, generating more efficient spatial plans with much less time," the study's abstract reads.
"Our method demonstrates the great potential of computational urban planning and paves the way for more explorations in leveraging computational methodologies to solve challenging real-world problems in urban science," it added.
The study's findings were recently published in the journal Nature Computational Science.
