Abstract: Embodiments described herein generally relate to a method and system for calibrating measured values for ambient air parameters using trained models. In at least one embodiment, the method includes calibrating measured values for an ambient parameter (AP) by generating training dataset points, each of the training dataset point comprising: (i) a first measured value of an AP, generated by a first sensor, at a time instance, and (ii) a second, time-paired measured value for the AP, generated by at least one reference sensor, at the time instance, wherein the first and second values are measured in a localized area surrounding the at least one reference sensor; training an AP-specific calibration model using the training dataset points to generate a trained AP-specific calibration model; applying the trained AP-specific calibration model to a new measured value of the AP, generated by a second sensor.

Abstract: Embodiments herein generally relate to methods and systems for using a forward calibration model to generate calibrated values for ambient parameters. In at least one example, the method for calibrating a measured un-calibrated ambient parameter (AP) value, includes generating, using an optimization model, a predicted calibrated AP value corresponding to the un-calibrated ambient parameter (AP); inputting, into a trained forward calibration model: (i) the predicted calibrated AP value; and (ii) one or more accuracy-enhancing parameters; generating, using the trained forward calibration model, a predicted un-calibrated AP value; determining if an error difference between the predicted un-calibrated AP value and measured un-calibrated ambient parameter (AP) value, is below a pre-determined threshold; if not, using the optimization model, to generate an updated predicted calibrated AP value and iterating the method, and otherwise outputting the predicted calibrated AP value, as the calibrated AP value.