Comparison of nonlinear filters for the estimation of parametrized spatial field by robotic sampling

The use of robotics in distributed monitoring applications requires wireless sensors that are deployed efficiently with an awareness of the information gain, communication constraints, resource allocation and coordination, and energy utilization. In this paper, we address the estimation of a parameterized spatial field distribution with a group of mobile robots sampling adaptively and using a statistically-aware algorithm. The proposed work investigates the use of different nonlinear filters, such as the Extended Kalman Filter (EKF) and some variants of it, and the Unscented Kalman Filter (UKF), both using adaptive sampling, so as to improve the speed and accuracy of the overall field distribution estimation scheme. The results from an extensive simulation work show that different variants of the standard EKF and the standard UKF can be used to improve the accuracy of field estimate and the main objective of this paper is to seek a practical trade-off between the desired field estimation accuracy and the computational load needed for this purpose.