INFORMATION TECHNOLOGY

AI Urban Spatial Planner is 3,000 times faster than humans


Li Yong, an associate professor in the Department of Electronic Engineering at Tsinghua University, and his collaborators have developed a machine learning model that can generate efficient land use and road planning for urban communities, surpassing other algorithms and human experts. The model shows how machine learning can be used to assist human planners in the complex task of spatial city planning. The research was recently published in Nature Computational Science.

Effective community spatial planning contributes to sustainable urban development. Urban spatial planning is often done by human designers after multiple rounds of analysis, discussion, and iteration. Especially for land use and the task of generating road plans, the potential solutions can be massive. This can hinder the process of urban planning, as human experts don’t have much time to focus on the more conceptual and creative steps of the process.

Li Yong and colleagues propose a model of a deep reinforcement learning algorithm that can perform complex urban spatial planning to generate optimal planning based on the concept of a 15-minute city (residents can walk or cycle to basic services in 15 minutes). Combined with human input, research shows that machine learning-assisted land and road spatial planning outperforms other algorithms and professional human designers, improving by about 50% and 3,000 times faster in all metrics considered.

The researchers also show how the method can be used to generate efficient plans for different planning strategies, such as situations where park and green space cover needs to be more important than other settings. This suggests that machine learning as a supporting tool increases the productivity of human planners, as well as the ability to potentially create efficient planning for more sustainable urban living.

Image from the author

(Source: China Science News Feng Lifei)

Related paper information:https://doi.org/10.1038/s43588-023-00503-5



Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button