Reducing multi-hop calibration errors in large-scale mobile sensor networks

摘要: Frequent sensor calibration is essential in networks with low-cost sensors. We exploit the fact that temporally and spatially close measurements of different sensors measuring same phenomenon are similar. Hence, when calibrating a sensor, we adjust its parameters to minimize differences between co-located previously calibrated In turn, freshly can now be used calibrate other network, referred as multi-hop calibration. first study respect reference signal (micro-calibration) detail. show ordinary least squares regression---commonly noisy sensors---suffers from significant error accumulation over multiple hops. this paper, propose novel algorithm using geometric mean regression, which (i) highly reduces propagation (ii) distinctly outperforms scenario, (iii) requires considerably fewer ground truth compared existing network algorithms. The proposed especially valuable large heterogeneous noise characteristics. provide theoretical justifications for our claims. Then, conduct detailed analysis artificial data accuracy under various settings identify sources. Finally, use accurately 13 million temperature, ozone (O3), carbon monoxide (CO) gathered by mobile air pollution monitoring network.