Quantifying sensing quality of crowd sensing networks with confidence interval

摘要

Quantifying sensing quality is fundamentally important for crowd sensing networks. Existing works which focus on quantifying the sensing quality of individual user are not applicable for that of the overall crowd sensing networks. However, it is nontrivial to quantify the sensing quality of crowd sensing networks for two main challenges. First, it is difficult to quantify the sensing quality of crowd sensing networks with the uncertainty, originating from the noisy sensing data and the complex inference of sensing information. Even worse, it is nearly impossible to conduct multiple times of repeated evaluation experiments to quantify the uncertain sensing quality based on the statistical information in crowd sensing networks, due to the dynamic nature of networks and the uncontrolled mobility of users. To address these challenges, inspired by the channel capacity, we investigate the physical essence of sensing quality deeply, and propose a confidence-interval based metric to quantify the uncertain sensing quality, leveraging the lower bound of sensing variance. Moreover, we exploit the Fisher Information of the crowd sensing data to compute the confidence interval without repeating evaluation experiments for multiple times. The extensive simulations demonstrate that our method can achieve a more accurate sensing quality for crowd sensing networks than the status quo methods.