Artificial Intelligence Systems Outperform Humans in Urban Planning Design

A team of urban planners and information scientists from Tsinghua University has found that AI-based urban planning systems can outperform human experts in designing city plans. In their study published in the journal "Nature Computational Science," the team describes the factors of an ideal city plan and their AI testing results. Paolo Santi from MIT's Senseable City Lab also published a commentary in the same journal, summarizing the team's achievements in this new work.

For most of history, cities have developed naturally, with planners and engineers trying to meet the determined needs. In many cases, this haphazard development has led to less-than-ideal results. In recent years, urban planners have sought to adopt more logical approaches to address this issue by creating design plans for developing areas that consider factors such as livability and pollution control.

This has led to the gradual development of urban planning as a science and the emergence of professional urban planners. With the increasing scale of new development projects, this science can become very complex as more variables need to be considered. In this new study, the research team found a way to apply AI to this problem to alleviate the burden on professionals.

When developing their AI system, the team chose to use the concept of 15 minutes, where residents of planned communities should be able to reach essential services within 15 minutes from their homes. Such a system can reduce travel time and air pollution caused by transportation, improving residents' quality of life. The researchers trained the system using previously developed plans by humans and factors considered to be good design features, such as parks, green spaces, bike lanes, and recreational areas.

Once the system was established, the research team first used it to design a small community in a 3x3 city block. This allowed them to fine-tune the system and make improvements. They then gradually increased the scale of development required and studied the results.

The researchers found that the plans developed by the system were as good as, if not better than, those created by humans. They also found that the time required was significantly reduced, from several hours to a few seconds. The team concluded that their system is not meant to replace human urban planners but rather to free them from the mundane aspects of city planning, allowing them to focus on the overall concept.