Orchestrating Participatory Sensing Campaigns With Workflows

Scriptie/masterproef: Master's Thesis


The rise of relatively cheap, Internet-connected, programmable, sensor-laden smart- phones has vastly increased the potential for person-centric applications. As a result, the idea of participatory sensing emerged, in which public and professional users participate in gathering, analysing and sharing local knowledge about different aspects of their environment. Especially in urban contexts, this form of sensing enables the assessment of environmental properties on a spatio-temporal scale and with a level of participation that was simply unattainable before.
Communities often want to tackle a local issue, for which participatory sensing can be used to gather relevant data. This data usually is gathered by running a campaign, as the absence of the crowd prevents participatory sensing to produce qualitative environmental data if the data is not collected in a focused effort. Orchestrating these campaign is a necessity, as letting the citizens handle a campaign themselves - without the domain knowledge and coordination - has the risk of not yielding qualitative results. This orchestration is usually handled by domain experts, which prevents citizens from running their own campaign autonomously. Additionally, this form of orchestration becomes unmanageable when campaigns increase in frequency and complexity. As a result, there is a need for automation.
In this thesis we investigate how we can enable citizens to create campaigns autonomously, by developing a framework that separates the domain expert knowledge from the local citizens' knowledge. This framework consist of two parts. First, the campaign definition framework, which enables citizens to autonomously create, monitor, raise awareness about, and contribute to participatory sensing campaigns. Second, the campaign orchestration framework, which automatically orchestrates campaigns, thus enabling citizens to complete a campaign successfully without the domain expert's knowledge. This orchestration framework is built around the notion of a workflow engine which orchestrates the participants, their contributions, and the overall campaign. We contribute a novel kind of workflow pattern specially designed for handling data flow.
While the concept of a campaign, and the issues of data density, are phrased in terms of participatory sensing in general, we evaluate our approach concretely in the context of environmental noise. In particular, we recreate a manually orchestrated noise mapping campaigns and show that our approach gives exactly the same results. Moreover, we show that our approach is able to do much more, by using dynamical analysis to coordinate participants and automatically recruiting additional participants that are implicitly contributing to a campaign.
Datum Prijs5 jul 2013
BegeleiderWolfgang De Meuter (Promotor), Ellie D'Hondt (Advisor) & Elisa Gonzalez Boix (Advisor)

Citeer dit