Imagine living in a green city full of parks, pedestrian walkways, bikeways, and buses that take people to stores, schools, and service centers within minutes.
This refreshing dream is an example of urban planning, which is summarized in the idea of the 15-minute city, where all basic needs and services are within reach in a quarter of an hour, leading to improved public health and reduced vehicle emissions.
This can be achieved with the help of artificial intelligence, as a new study conducted by researchers at Tsinghua University in China revealed how machine learning can generate spatial layouts more efficiently than humans, and in record time.
Automation scientist Yu Cheng and the research team developed an artificial intelligence system to handle the computational tasks of urban planning, and found that it produces urban plans that outperform human designs by about 50% on three metrics: access to services, green spaces, and traffic levels.
The team started with a small project that involved modeling urban areas with an area of just a few square kilometres.
After two days of training, and using several neural networks, the AI system searched for optimal road and land use layouts that fit the 15-minute city concept and local planning policies and needs.
While the AI model developed by Cheng and his colleagues has some features to extend its use to planning larger urban areas, designing an entire city would be infinitely more complex.
But automating some steps in the planning process can save a huge amount of time: the AI model calculated certain tasks in seconds, while it took human planners 50 to 100 minutes to complete them.
The researchers say that automating the most time-consuming tasks in urban planning would free planners to focus on more challenging or human-centered tasks, such as public engagement and aesthetics.
The research team envisions the AI system acting as an “assistant” for urban planners, who can create designs that are optimized by algorithms, and reviewed, modified and evaluated by human experts based on community feedback.
The study was published in the journal Nature Computational Science.
Source: ScienceAlert - Publication date: 09/21/2023 - https://ar.rt.com/w2du