Figure 1.
RFID sensor system and system deployments.
A) Schematic illustration of the RFID sensor system. RFID tags are worn as badges by the individuals participating to the deployments. A face-to-face contact is detected when two persons are close and facing each other. The interaction signal is then sent to the antenna. B)C)D) Activity pattern measured in terms of the number of tagged individuals as a function of time in the three deployments: B) ISI refers to the deployment in the offices of the ISI foundation in Turin, Italy, with 25 participants; C) 25C3 to the 25th Chaos Communication Congress in Berlin, Germany, with 575 participants, and D) SFHH to the congress of the Société Française d'Hygiène Hospitalière, Nice, France, with 405 participants. Dashed vertical lines indicate the beginning and end of each day. Typical daily rhythms are observed in the office and conference settings. The ISI deployment allows us to recover the weekly pattern signaled by the absence of activity on the day of Sunday (the number of persons larger than zero at night indicates the tags left in the offices, easily recognizable from the flat behavior).
Figure 2.
Probability distribution of human interactions.
A) Probability distribution of duration of contacts between any two given persons. Strikingly, the distributions show a similar long-tail behavior independently of the setting or context where the experiment took place or the detection range considered. The data correspond to respectively 8700, 17000 and 600000 contact events registered at the ISI, SFHH and 25C3 deployments. B) Probability distribution of the duration of a triangle. The number of triangles registered are 89, 1700 and 600000 for the ISI, SFHH and 25C3 deployments. C) Probability distribution of the time intervals between the beginning of consecutive contacts AB and AC. Some distributions show spikes (i.e., characteristic timescales) in addition to the broad tail; for instance, the 1 h spike in the 25C3 data may be related to a time structure to fix appointments for discussions.
Figure 3.
A) Distribution of contact durations (in seconds) at the 25C3 deployment, for various time intervals and for the entire dataset. The filled symbols correspond to the distribution of contact durations of several individual tags. B) Distribution of contact durations (in seconds) for sampled datasets in which 60% of the tags are ignored, compared with the distributions obtained from the whole datasets.
Figure 4.
Properties of the aggregated network of contacts corresponding to the third 12-hour period of the 25C3 deployment. The total number of packets exchanged by a tag during a contact (strength s) is shown as a function of the number of distinct contacts (degree k). A superlinear (powerlaw) behavior is observed, with a slope of 1.73 [95%CI: 1.65–1.81] obtained from the fitting procedure with a correlation coefficient of 0.93. Inset: distribution of links' weights, defined as the total number of packets exchanged between two interacting tags. The same qualitative properties are obtained for other time intervals and for all the other experiments we deployed.
Figure 5.
From RFID communications to contact networks.
Top: Temporal aggregation of proximity relations reported by different tags over a sliding window. The information collected by each tag is aggregated and translated into a dynamical adjacency matrix to reconstruct the dynamical network of face-to-face interactions. Bottom: Real-time visualization. A snapshot of the visualization, displaying approximate position information as well as the instantaneous network of face-to-face proximity. Individuals wearing an RFID tag are represented as discs labeled with the numeric identifier of their tag. Edges between individuals represent ongoing face-to-face proximity relations, and their thickness reflects the strength of the proximity relations. The other labels refer to names of rooms in the venue and denote the location of RFID readers. The graph is laid out so that individuals are shown near the readers that report their presence, and the sizes of the readers symbols depend on the number of users from which they receive information.