Q-Adaptation of UKF Algorithm for Estimation of the Autonomous Underwater Vehicles Dynamics

Abstract

Accurate estimation for the pose and orientation of the autonomous underwater vehicles (AUVs) is necessary in most of the applications. This is same for both the missile type AUVs or the vehicles used for research purposes deep in the sea or ocean. A nonlinear version of the Kalman Filter as the unscented Kalman filter (UKF) gives satisfactory estimation results for this purpose in the normal operation conditions. However, in the deep sea, changes in the environment (process noise) either instantaneously or periodically are very likely. In such case, the UKF must be adapted to become robust against such changes and provide accurate estimation results even in this case. In this study, we propose process noise covariance matrix adaptation (Q-adaptation) for the UKF algorithm. The main aim is to make the algorithm adaptive against the changes in the process noise covariance. The Adaptive UKF (AUKF) estimates the AUVs dynamics. We investigate the performance of the algorithm when the process noise increases and decreases, which are very likely as a result of changes in the AUV dynamics in different environments. The results are compared with those of a non-Q-adaptive version of the UKF algorithm.