Application of Polynomial Models in Bayesian Fusion of Humidity Sensors

Abstract

Local polynomial trend models are a special class of state-space models that can be used without having the full information about the process under study, since most of their parameters are embodied in the state vector and estimated immediately. This makes them attractive for use in signal processing. The present work considers problems that arise when using a polynomial model with a local quadratic trend for Bayesian fusion of two humidity sensors. The unknown sensor biases make it impossible for the model to satisfy the observability conditions. There is currently no general solution to this problem. To overcome this difficulty, an approach is presented where the humidity measurement result implicitly includes the bias of one of the sensors. The results of the study can be used to fuse quantities other than humidity when two or more sensors are available.