AI-generated image of a legal city grid.

Using artificial intelligence to design cities

To meet the challenges of climate change, sustainability, and more efficient living, cities will need to be better designed than they were in the twentieth century.y a century. We have generally learned what features make a city or urban space more useful to us and more adaptable to the environment; However, the planning of such communities still depends largely on human input and expertise. This work also needs to navigate complex geographies constrained by existing urban features or particular settings.

Using artificial intelligence to design sustainable urban areas

Researchers are now trying to use artificial intelligence techniques, including deep reinforcement learning, to design more efficient and sustainable urban spaces despite these limitations and challenges.

Previously, scientists developed computational methods to improve urban plans. However, these plans are not always acceptable to urban stakeholders even if they are more efficient.

One solution might be to combine human and computer decision-making, enabling more acceptable and sustainable urban plans based on accepted metrics. Deep learning, which uses artificial neural networks to solve tasks, has become popular in recent years; Urban planning has used deep learning in ways to solve specific problems, such as transportation.

Can artificial intelligence be used to improve city planning? Image generated by artificial intelligence.

Can artificial intelligence be used to improve urban planning?

However, deep learning methods have not been widely used to map spatial layouts of multiple urban elements or even entire cities.[1] What makes the problem particularly difficult is that urban areas generally have irregular layouts due to restricted geographic areas or outdated urban designs that limit any new construction.

To solve this problem, one solution is to make a series of tasks and decisions necessary to create more acceptable and sustainable layouts. This is done through a Markov process that is then solved using reinforcement learning, an artificial intelligence technique that optimizes specific choices (for example, the decision to build a new road) that lead to better rewards or outcomes.

In this approach, urban areas are characterized as an urban connectivity graph (UCG), where place elements, including roads or land parcels, are expressed as topological and spatial relationships.

When a new element, representing a new layout element such as a new road, is added to the topological relationship expressed in the graph, the layout is evaluated anew. This can be represented by the workspace produced by a Markov process. In this case, the graph evaluates three main elements, including access to services, access to parks, and the efficiency of the road network.

An AI drawing of a city with buildings organized in circles.
Research has shown that the use of deep learning improves planning solutions in the areas of park access, traffic and service access. Generative Image Artificial Intelligence.

However, the problem cannot be solved easily, as spatial heterogeneity can make this problem difficult and many options exist on how to achieve planning for the urban environment. To make the problem more feasible, a graph neural network (GNN) is used to express geographic data as a graph so that the problem becomes constrained. Planning situations are evaluated through graphical relationships and choices are compared to other possibilities.

Business and value grids are created, with value grids used to evaluate bidding actions for proposed layouts. The information in UCG is evaluated using a GNN that produces certain results. Overall, this approach results in better results in the three categories evaluated, namely improved access to services, environmental use (i.e. access to parks), and better traffic flow, when compared to human-created planning by about 50%.[2],[3]

Combining artificial intelligence and human input for urban planning

Although the results show that an AI-based approach can improve urban planning better than human-only planning, relying on algorithmic planning may not be the best option. As with other computational methods, sometimes the most efficient method may not be acceptable to communities.

Instead, a combination of human and computer planning may be a better approach. For example, urban plans and layouts can be fed into a deep reinforcement learning approach, with urban planning results examined to see if they are acceptable to stakeholders or if adjustments are needed.

The benefit of this approach is that it not only allows plans to be more efficient and acceptable to communities, but can reduce the planning process by a factor of up to 3,000. This method is also more efficient than other computing methods for planning, while it enables planners to focus on visualizing Create a new plan and then evaluate its final results. Overall, design experiments show that this approach is 12% and 5% more efficient in performance metrics and reduces overall planning time.

However, the challenge of using AI to design cities has not been solved. Even by making this approach more efficient using UCGs and GNNs, designing a large city becomes extremely more complex with space metrics. Not only does the size of cities become an issue for any massive graphics computing requirements, but large cities are also complex systems containing many social and urban environmental layers that may make solving tasks more difficult, at least in a way that may be acceptable to local communities and many owners. interest.

For now, local use of deep reinforcement learning within neighborhoods or small communities may be the best option.


[1] An example of deep learning for transportation planning can be found here: Aqeeb M, Mahmoud R, Al-Zahrani A, et al. (2019) Rapid transit systems: smarter urban planning using big data, in-memory computing, deep learning, and graphics processing units. Sustainability 11(10): 2736.

[2] The article discussing the use of deep reinforcement learning and its application to urban design can be found here: Zheng Y, Lin Y, Zhao L, et al. (2023) Spatial Planning of Urban Communities via Deep Reinforcement Learning. Computational natural sciences. Epub ahead of print on Sep 11, 2023. DOI: 10.1038/s43588-023-00503-5.

[3] An article that helps explain key research by Zheng et al. It can be found here: Santi P (2023) Artificial intelligence improves the design of urban communities. Computational natural sciences. Epub ahead of print on Sep 11, 2023. DOI: 10.1038/s43588-023-00515-1.



Leave a Reply

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