Fig. 2. Boston metropolitan area. Blue circles represent the starting locations of emergency alarms (with 100 m radius). (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)
The main difference between our work and the cited works lies in the traced participants selection. Our approach does not
rely on volunteers who agreed to be tracked, who usually are students or conference participants who move in a small scale
scenario. We instead use data that is generated by anyone using their mobile phones and connected to a particular telecom
operator. This allows us to indirectly monitor the human mobility of a wider range of people and over larger temporal and
spatial intervals.
Inferred traces have been considered in [10] using urban transport data, and routine behavior has been exploited for
media sharing. Mobile phone data were considered in [11] to infer contact events between roughly two thousand people
in the Boston metropolitan area. However, none of the studies have considered large gathering events. Moreover, none of
the works presented above have studied the impact of human mobility patterns for enabling emergency communications
when traditional communication infrastructures fail.
In this paper, we instead evaluate the feasibility of human mobility: (i) at large temporal and spatial scale; (ii) not relying
on volunteers who agreed to be tracked; (iii) to disseminate emergency information in absence of network infrastructures
and in a timely fashion. To the best of our knowledge, this is the first work in which a real-world wide-scale data set has
been considered to enable emergency communications thorough a partially-connected ad hoc network.