상세정보

Bluetooth Low Energy (BLE) is probably the best technological tool that we can harness today for studies on close human interactions. It has a wide deployment base (i.e., on smartphones), has peer discovery as an inherent protocol feature, does not require infrastructure (e.g. satellites or base stations) to operate, and sparingly uses energy that is good for extended monitoring. In this dissertation, we show that we can use the BLE peer discovery capability on smartphones to detect and monitor massive and dynamic encounters, which would provide valuable insights into many epidemiological or sociological phenomena. However, being designed for more leisurely interactions, BLE needs some stretching in order to be used in large-scale operations. Specifically, the protocol design is not optimal to rapidly discover hundreds of devices in the communication range, whereas dense crowds and mass gatherings are not unrealistic in city life. Moreover, if the crowd is dynamic, discovery becomes even more time-pressed because encounters should be recorded before churn. In this dissertation, we push the BLE technology with the requirements to discover hundreds of devices before co-presence expires, and to work continually over a typical smartphone charge cycle. Specifically, we investigate how we should modify the BLE protocol and how we should set its protocol parameters for this purpose. We show that with the proposed changes and configurations, we can accelerate the speed of discovery for massive and dynamic crowds by more than an order of magnitude compared to the case that we naively follow the guidance of the current BLE standard.

1 Introduction 1
 1.1 Organization of the Dissertation  6
2 Related Work 7
 2.1 GPS. 8
 2.2 RFID and sensor network 9
 2.3 Social media and search record  10
 2.4 Wi-Fi 11
 2.5 Bluetooth 12
 2.6 Energy-Ecient Discovery Protocol 14
 2.7 Bluetooth LE Analysis . 16
3 Encounter Monitoring with BLE 18
 3.1 System design 19
 3.2 Implementation  25
4 Speed of Discovery 28
 4.1 A Game of Two Random Walks  29
 4.2 Spacing ADV PDUs and Scans  38
 4.3 A Base Case Analysis  41
 4.4 System Dynamics in T, S, and m 45
  4.4.1 Impact of m . 45
  4.4.2 Impact of S . 49
  4.4.3 Impact of T . 50
 4.5 Robustness Under Packet Losses 53
  4.5.1 Real-life Packet Loss Rate 53
  4.5.2 Analysis 55
  4.5.3 Spacing of advertisements revisited .59
 4.6 Summary  60
5 Energy consumption 61
 5.1 Simulation Setup  62
 5.2 Results 65
6 Population Estimation for Adaptation 68
7 Conclusion 73
Bibliography 75