On-line sensor calibration and error modeling using single actuator stimulus

We have developed an on-line in-field nonparametric calibration and error modeling approach. The approach employs a single excitation source as the external stimulus to create differential sensor readings. Under very mild assumptions imposed on the calibration functions, error model and the environment model, the technique utilizes the maximal likelihood principle and a nonlinear function minimization to derive both simultaneously the calibration function and the error model of a specified accuracy. Resubstitution is then used in order to establish the interval of confidence. The approach is intrinsically localized and we present two variants: i) one where only pairs of neighboring sensors communicate in order to conduct calibration and construct error model; ii) one where a provably minimum amount of communication is achieved. While the idea of employing external actuators to conduct calibration is generic in the sense that it can be applied to any sensor modality, in this paper we demonstrate and evaluate the approach using traces from light sensors and acoustic signal-based distance measurements recorded by in-field deployed sensors.