The image of the city through the eyes of machine reasoning

The majority of humans will live in cities by the end of 2050, and according to Forbes magazine, hundreds of new cities will be built in China alone in the coming decade. City layouts have an important impact on many aspects of city life, e.g., transportation, urban dynamics, urban morphology, and the quality of life. There is a deep correlation between urban form and urban performance and it is important to consider urban layouts in terms of their effects on the quality of life in cities. This chapter investigates the development of a machine reasoning (MR) engine that learns common design patterns from the layout of a small number of reference cities. In this study, we selected six cities as positive and negative examples based on their rank in a quality of life index. The MR identifies latent patterns within these city layouts based on a set of policies and builds a vocabulary starting with lower-level urban patterns that are combined hierarchically to develop higher-level concepts corresponding to more complex city patterns. Our approach differs significantly from the current neural network-based approaches that require big data. Instead, through the MR tool we identify seminal patterns within cities with limited data. Our approach has the potential to open up research in fields, where data are scarce, e.g., there are a limited number of city layouts in the world. This research has the potential to automate and guide city planners to develop high-performing city layouts in the future.