My previous paper/post on distributed expectation-based spatio-temporal clustering algorithms discussed spatio-temporal cluster detection in a distributed system and how clusters can be used for efficient opportunistic message delivery in Pocket Switched Networks (PSNs).
Our newest paper to appear in the Journal of Ad Hoc Networks takes this idea one step further and defines a different algorithm with which spatio-temporal clusters can be detected in PSNs. The new Distributed Rise And Fall spatio-Temporal (DRAFT) clustering algorithm is a non-monotonic distributed clustering algorithm and message delivery protocol for PSNs. DRAFT delivers more messages more efficiently than the current state of the art protocols, including Bubble and PRoPHETv2.
The paper also contains some new insights into how humans congregated into transient groups. So if you're interested in human behavioural analysis then the paper should have something for you.
If you'd like to try out DRAFT for yourself you can download the DRAFT and DEBT code bundle for The ONE simulator with example config files from this link.