Blind Calibration of Networks of Sensors: Theory and Algorithms

With the wide variety of sensor network applications being envisioned and implemented, it is clear that in certain situations the applications need more accurate measurements than uncalibrated, low-cost sensors provide. Arguably, calibration errors are one of the major obstacles to the practical use of sensor networks [3], because they allow a user to infer a difference between the readings of two spatially separated sensors when in fact that difference may be due in part to miscalibration. Consequently, automatic methods for jointly calibrating sensor networks in the field, without dependence on controlled stimuli or high-fidelity groundtruth data, is of significant interest. We call this problem blind calibration.