To realize the ultimate vision of Internet of Things (IoT), only depending on terrestrial network is far from enough. As a supplement and extension of terrestrial network, satellite network can offer powerful support to realize the depth and breadth of the coverage. However, existing satellite networks are usually designed for particular purposes. Moreover, traditional satellite networks and terrestrial networks are developed and operated separately, consequently they cannot meet the need of network flexibility required by IoT. In this paper, a space-terrestrial architecture is conceived for constructing a space-terrestrial based IoT (ST-IoT) system. Additionally, a reliable identification procedure, an integrated access and communication procedure, as well as a clustering cooperative transmission strategy are also presented.

Human-centric service is an important domain in smart city and includes rich applications that help residents with shopping, dining, transportation, entertainment, and other daily activities. These applications have generated a massive amount of hierarchical data with different schemas. In order to manage and analyze the city-wide and cross-application data in a unified way, data schema integration is necessary. However, data from human-centric services has some distinct characteristics, such as lack of support for semantic matching, large number of schemas, and incompleteness of schema element labels. These make the schema integration difficult using existing approaches. We propose a novel framework for the data schema integration of the human-centric services in smart city. The framework uses both schema metadata and instance data to do schema matching, and introduces human intervention based on a similarity entropy criteria to balance precision and efficiency. Moreover, the framework works in an incremental manner to reduce computation workload. We conduct an experiment with real-world dataset collected from multiple estate sale application systems. The results show that our approach can produce high-quality mediated schema with relatively less human interventions compared to the baseline method.