From physical sensors to human-as-a-sensor: human behavior learning with heterogeneous cyber-physical systems
Description
TitleFrom physical sensors to human-as-a-sensor: human behavior learning with heterogeneous cyber-physical systems
Date Created2021
Other Date2021-10 (degree)
Extent1 online resource (xiv, 130 pages) : illustrations
DescriptionSince the Coke machine was first connected on the Internet in Pittsburgh, an idea be- gan to spread: a vision of an interconnected world where information on most everyday objects was accessible. Following this vision, scientists and engineers have developed this idea into a concept called Cyber-Physical Systems (CPS) or Internet of Things (IoT), which interconnects a large number of heterogeneous devices, from traditional desktop PCs to smartphones, from household appliances to wearable devices. However, a key element, e.g., human beings, is less studied compared to the physical and cyber components, which is of great importance considering the intrinsic purpose of CPS is to serve humans.
In this dissertation, we study the problem of human behavior learning and decision making in heterogeneous Cyber-Physical Systems. Our work explores a wide range of large-scale urban physical systems including a citywide cellular network system, a citywide on-demand delivery system, a statewide payment system, and a nationwide vehicular system. Based on the data collected from these systems, our work focuses on two fundamental challenges that are unique in the human-in-the-loop urban CPS. Based on the data collected from these systems, my work focuses on two fundamental challenges that are unique in the human-in-the-loop urban CPS: (i) Human Behavior Learning: learning human behavior such as vehicular mobility behavior, data usage behavior, payment behavior, and delivery behavior; (ii) Decision Making: making decisions based on the knowledge from human behavior learning such as traffic control, load balancing, insurance pricing, and delivery scheduling. Specifically, for human behavior learning, we focus on three aspects: (i) learning from stationary sensing systems: we present a real-time vehicular mobility prediction model VeMo, which predicts the locations of all the vehicles on highways only based on extreme sparse data collected from electronic toll collection systems; (ii) learning from mobile sensing systems: we present a prediction model Mohen based on two heterogeneous systems to complement each other, which predicts vehicular mobility even after leaving the coverage of one. (iii) learning from human-as-a-sensor sensing systems: we present an indoor localization method to infer couriers’ indoor mobility based on couriers’ manual reporting in an on-demand delivery platform. These three works are implemented and evaluated with real-world data from two Chinese cities Shenzhen and Shanghai. The results demonstrate human behavior learning and decision making can be well integrated into urban CPS, which outperform the state-of-the-art in different scenarios.
NotePh.D.
NoteIncludes bibliographical references
Genretheses
LanguageEnglish
CollectionSchool of Graduate Studies Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Statistical Profile