Before you ever step into a new coffee shop, years of urban planning have gone into building that brick-and-mortar building in your neighborhood.
Urban planning is a multi-faceted discipline in which planners collect data points from targeted areas to determine if a store, hospital or other types of buildings are likely needed in that area. It also involves determining if the needed infrastructure to support the building is there or needs to be added. The data provides context to planners and includes things such as geographic factors, socioeconomic statistics and human mobility.
This kind of planning can take years to configure, but for first-year computer science doctoral student Dongjie Wang, that isn’t efficient. Wang has developed an artificial intelligence program called the Land-Use Configuration Generative Adversarial Network (LUCGAN). This program is made to learn the context of an area through human data input and then develop solutions for urban planners. It could potentially save time and money for communities involved in planning new neighborhoods.
LUCGAN is based on a deep Generative Adversarial Network, a machine-learning technique in which a machine is trained in a specific ‘set.’ This learning occurs after the machine is presented with a data distribution, or a list that includes all points of data; with this information the machine can produce infinite probability points that could be inserted into a graph, or in this case an axis.
Deep learning is patterned in many ways after human learning, but where the brain connects learning through neurons, LUCGAN uses networks to gather data points and correlate them.
LUCGAN uses networks to undergo training after the context area data is uploaded into the program. Thus, the AI is able to produce the most efficient “probability point,” which in urban planning is the area that could most efficiently support the new development.
Wang’s idea could save not only time but could also prevent human error and save money. When urban planning is done badly, entire communities end up paying for a planner’s failure to consider needed infrastructure and services.
When these issues are not accounted for, residents end up paying for unnecessary services such as new bus routes or extra gas money for the commute. If a neighborhood is built without a nearby fire station, hospital, or even grocery store sometimes taxpayers end up having to pay. With the prospect of eliminating human error, LUCGAN could play a powerful role in reducing these mistakes by taking all land-use context points into consideration without human error.
Wang and his advisor, Professor Yanjie Fu, are currently in the process of disseminating this open-source software publicly to experts in the field with the hope of improving and shifting the urban planning model.
“Artificial intelligence is used in so many areas in the modern world,” says Wang, “why not urban planning?”