Effective Urban Region Representation Learning Using Heterogeneous Urban Graph Attention Network (HUGAT) 

Revealing the hidden patterns shaping the urban environment is essential to understand its dynamics and to make cities smarter. Recent studies have demonstrated that learning the representations of urban regions can be an effective strategy to uncover the intrinsic characteristics of urban areas. However, existing studies lack in incorporating diversity in urban data sources. In this work, we propose heterogeneous urban graph attention network (HUGAT), which incorporates heterogeneity of diverse urban datasets. In HUGAT, heterogeneous urban graph (HUG) incorporates both the geo-spatial and temporal people movement variations in a single graph structure. Given a HUG, a set of meta-paths are designed to capture the rich urban semantics as composite relations between nodes. Region embedding is carried out using heterogeneous graph attention network (HAN). HUGAT is designed to consider multiple learning objectives of city’s geo-spatial and mobility variations simultaneously. In our extensive experiments on NYC data, HUGAT outperformed all the state-of-the-art models. Moreover, it demonstrated a robust generalization capability across the various prediction tasks of crime, average personal income, and bike flow as well as the spatial clustering task. 

Regionalization for urban air mobility application in metropolitan areas: case studies in San Francisco and New York

In this study, as a first step to assess the feasibility of UAM in urban areas, we conduct 3D geodemographic analyses of two major cities in San Francisco, CA and Manhattan, NY. The 3D building footprint data is used to identify the raw available airspace as well as the added spatial restrictions with geofencing. Population data is used to represent the potential customer base by combining the daytime and nighttime population. Since the geospatial and demographic datasets differ in representation, the spatial data is vectorized while population data is available by census tract, spatial information is aggregated in census tracts. In addition, We proposed to group the areas of similar spatial and population characteristics through regionalization. Regionalization is a spatially constrained multi-variate clustering method to group small geographical units (census blocks and tracts in general) into a contiguous region of homogeneous nature. The main benefit of regionalization is to delineate regions of similar characteristics and spatial proximity. Through regionalization, one can better understand the urban space with comprehensive geographic perspective, rather than a small geographical unit of census blocks or tracts. Furthermore, regionalization can also improve geospatial intelligence in urban spaces by delineating the functional neighborhood . In this study, we adopted the SKATER, an efficient regionalization technique that uses minimum spanning tree consisting of a connected tree with no circuits. The intention is to provide a region map of the city that can readily identify regions of similar UAM operational and population characteristics with spatial continuity and feasibility. Based on the regionalization results, correspondence analysis was conducted to translate the compound effect of spatial and population characteristics into feasibility