TY - JOUR
T1 - Crowdsourcing emergency data in non-operational cellular networks
AU - Chatzimilioudis, Georgios
AU - Costa, Constantinos
AU - Zeinalipour-Yazti, Demetrios
AU - Lee, Wang Chien
N1 - Funding Information:
This work was supported by an Appcampus Award (Microsoft, Nokia and Aalto University, Finland ). It has also been supported by the third author׳s startup grant at the Univ. of Cyprus, EU׳s FP7 “Mobility, Data Mining, and Privacy” project, EU׳s COST Action MOVE “Knowledge Discovery for Moving Objects” and an industrial award by MTN Cyprus.
Publisher Copyright:
© 2015 Elsevier Ltd
PY - 2017/3/1
Y1 - 2017/3/1
N2 - In overloaded or partially broken (i.e., non-operational) cellular networks, it is imperative to enable communication within the crowd to allow the management of emergency and crisis situations. To this end, a variety of emerging short-range communication technologies available on smartphones, such as, Wi-Fi Direct, 3G/LTE direct or Bluetooth/BLE, are able to enable users nowadays to shape point-to-point communication among them. These technologies, however, do not support the formation of overlay networks that can be used to gather and transmit emergency response state (e.g., transfer the location of trapped people to nearby people or the emergency response guard). In this paper, we develop techniques that generate the k-Nearest-Neighbor (kNN) overlay graph of an arbitrary crowd that interconnects over some short-range communication technology. Enabling a kNN overlay graph allows the crowd to connect to its geographically closest peers, those that can physically interact with the user and respond to an emergency crowdsourcing task, such as seeing/sensing similar things as the user (e.g., collect videos and photos). It further allows for intelligent synthesis and mining of heterogeneous data based on the computed kNN graph of the crowd to extract valuable real-time information. We particularly present two efficient algorithms, namely Akin+ and Prox+, which are optimized to work on a resource-limited mobile device. We use Rayzit, a real-world crowd messaging framework we develop, as an example that operates on a kNN graph to motivate and evaluate our work. We use mobility traces collected from three sources for evaluation. The results show that Akin+ and Prox+ significantly outperform existing algorithms in efficiency, even under a skewed distribution of users.
AB - In overloaded or partially broken (i.e., non-operational) cellular networks, it is imperative to enable communication within the crowd to allow the management of emergency and crisis situations. To this end, a variety of emerging short-range communication technologies available on smartphones, such as, Wi-Fi Direct, 3G/LTE direct or Bluetooth/BLE, are able to enable users nowadays to shape point-to-point communication among them. These technologies, however, do not support the formation of overlay networks that can be used to gather and transmit emergency response state (e.g., transfer the location of trapped people to nearby people or the emergency response guard). In this paper, we develop techniques that generate the k-Nearest-Neighbor (kNN) overlay graph of an arbitrary crowd that interconnects over some short-range communication technology. Enabling a kNN overlay graph allows the crowd to connect to its geographically closest peers, those that can physically interact with the user and respond to an emergency crowdsourcing task, such as seeing/sensing similar things as the user (e.g., collect videos and photos). It further allows for intelligent synthesis and mining of heterogeneous data based on the computed kNN graph of the crowd to extract valuable real-time information. We particularly present two efficient algorithms, namely Akin+ and Prox+, which are optimized to work on a resource-limited mobile device. We use Rayzit, a real-world crowd messaging framework we develop, as an example that operates on a kNN graph to motivate and evaluate our work. We use mobility traces collected from three sources for evaluation. The results show that Akin+ and Prox+ significantly outperform existing algorithms in efficiency, even under a skewed distribution of users.
UR - http://www.scopus.com/inward/record.url?scp=84950349886&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84950349886&partnerID=8YFLogxK
U2 - 10.1016/j.is.2015.11.004
DO - 10.1016/j.is.2015.11.004
M3 - Article
AN - SCOPUS:84950349886
SN - 0306-4379
VL - 64
SP - 292
EP - 302
JO - Information Systems
JF - Information Systems
ER -