Nonlinear Filtering Approaches to Field Mapping by Sampling Using Mobile Sensors

This work proposes a novel application of existing powerful nonlinear filters, such as the standard Extended Kalman Filter (EKF), some of its variants and the standard Unscented Kalman Filter (UKF), to the estimation of a continuous spatio-temporal field that is spread over a wide area, and hence represented by a large number of parameters when parameterized. We couple these filters with the powerful scheme of adaptive sampling performed by a single mobile sensor, and investigate their performances with a view to significantly improving the speed and accuracy of the overall field estimation. An extensive simulation work was carried out to show that different variants of the standard EKF and the standard UKF can be used to improve the accuracy of the field estimate. This paper also aims to provide some guideline for the user of these filters in reaching a practical trade-off between the desired field estimation accuracy and the required computational load.