A hierarchical indoor and outdoor model for semantic trajectories

Abstract

With the growth of location-based information and the widespread adoption of mobile devices and connected sensors, human mobility has recently emerged as an important research area. Nowadays, the exponential development of mobile sensors and the Internet of Things offers many opportunities for the integration of real-time data on humans acting in indoor and outdoor environments. Moreover, mobile crowd-sensing allows volunteers to actively provide real-time trajectory and activity data (Guo et al., 2015). However, such crowd-sourcing data are most often heterogeneous in space and time and require a flexible data model that can integrate the data as they are and provide data manipulation and analysis capabilities to reformat the data at the appropriate level of abstraction. Understanding urban mobility patterns, together with associated contextual information, requires a sound integration of the modelling level within current information infrastructures. Such development appears as

A deep pedestrian trajectory generator for complex indoor environments

“…There have been some discussions on the rules that indoor trajectories should follow (Kontarinis et al., 2021; Noureddine et al., 2022). Based on the literature and common sense, we consider the following four groups of semantic rules for the current state in different scenarios: (1) starting to walk: the next feature node must be one of the entrances to the walkable space; (2) completing a certain activity: the next feature node should not repeat the previous activities; (3) terminating walk: the next feature node must be an exit; and (4) being located on a particular floor: the next feature node should be located on the same floor or an elevator.…”

“…In recent years, studies on indoor location‐based services (LBS), such as indoor service recommendation, indoor space modeling, walking pattern discovery, route prediction, and agent‐based epidemic simulation, have proliferated (Chen et al., 2019; D'Orazio et al., 2020; Guo et al., 2016; Harweg et al., 2021; Kontarinis et al., 2021; Mao & Li, 2020; Noureddine et al., 2022; Wang et al., 2019, 2022; Xiao et al., 2021). To provide LBS in venues such as shopping malls, exhibition centers, and conference halls, service providers need to model the movement behavior of pedestrians, which are expected to have the following characteristics: (1) randomness: pedestrian movements are influenced by a variety of factors, and their trajectories usually exhibit randomness with no apparent premeditated purpose; (2) relatively long duration: wandering movements may last for hours and cover large indoor areas (e.g., up to thousands of square meters); (3) rich semantics: indoor pedestrian trajectories are usually associated with rich semantics since pedestrians are engaged in a variety of activities such as shopping, waiting, or attending conferences; and (4) strict topological constraints: pedestrian movements must conform to the topological constraints of indoor environments.…”

STO2Vec: A Multiscale Spatio-Temporal Object Representation Method for Association Analysis

“…There are currently numerous studies analyzing the association relationships based on spatio-temporal co-occurrence from different perspectives. These studies can be classified into semantic trajectory-based approaches [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] and location embedding-based approaches [23][24][25][26][27][28][29][30][31][32][33] according to the analysis methods.…”

“…Associations are universal [3], making association analysis applicable to various fields with different relationship types. Sharma et al [2] grouped spatio-temporal associations into three types based on whether a temporal sequence was considered: sequential (e.g., analyzing event-oriented spatio-temporal association in video surveillance [4]), cascading (e.g., studying relationships between events, locations, and criminal activities in criminal geography [5]), and co-occurrences (e.g., similar associations between trajectories [6], co-location patterns between geographic entities [7], semantic annotation of trajectories [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22], and location embedding [23][24][25][26][27][28][29][30][31][32][33], etc.). By comparing their frequency of co-occurrence, spatio-temporal co-occurrence-based association analysis can reveal implicit associations between entities.…”

A Semantic Model of Internet of Things for Intelligent Translation and Learning

“…Examples are mainly used to describe the specific things contained in a certain class of objects. Relationship is used to represent the ownership relationship between things and attributes [16]. Property is used to describe the properties of objects and sub-objects.…”