PassFit: Participatory Sensing and Filtering for Identifying Truthful Urban Pollution Sources

With the increasing ubiquitous usage of the smart phone, participatory sensing networks become applicable for urban sensing applications. Notably, in pollution monitoring applications, untrained participants and heterogeneously unreliable sensors bring much noise into the reported data. Status quo researches fail to identify truthful pollution sources when the parameters are not known. In this paper, we cluster the reported data and make parameter estimations of the pollution sources. Inspired by the expectation maximization method, we estimate the existence of the pollution sources and leverage these results for noise estimations iteratively. The key insight is that, we can identify the truthful pollution sources with the noisy data mitigation iteratively. Further, theoretical analysis also shows that, we can achieve optimal estimation in the sense of maximum likelihood (ML). Simulation results show that, compared with basic ML algorithm, we improve the false positive and false negative of identification by 99% and 82%, respectively, at the same time, the mean error and the deviation error of noise estimation by 49% and 70%, respectively.