Previous Abstract Return to Session C6 Next Abstract

Session C6: Collaborative and Networked Navigation

Improved Maximum Correntropy Cubature Kalman Filter for Cooperative Localization
Shengxin Li, Bo Xu, and Asghar A. Razzaqi, Harbin Engineering University, China
Location: Atrium Ballroom
Alternate Number 1

Estimating accurate underwater position information is still one of the main technical challenges for Autonomous Underwater Vehicle (AUV). The positioning system must provide accurate positioning information over long distances and long periods of time, especially when the AUV is in some special environment; such that it cannot get GPS information from outside, and it cannot correct its position information by means of terrain matching or gravity. The high-precision inertial navigation system can provide high accuracy for short-time navigation; however, for the long-time navigation it will accumulate sufficient error over time, resulting in low positioning accuracy. The navigation method based on underwater acoustic such as long baseline (LBL), short baseline (SBL) and ultra-short (USBL) methods have high accuracy. However, underwater acoustic navigation methods need to install multiple hydrophones on the carrier and deploy arrays on the underwater and the technical operation is complex, which greatly limits the operating range of the AUV. In recent years, the sonar technology is used to measure the relative distance information between AUVs, and the Cooperative Localization (CL) of multiple AUVs is realized, which has become a key research issue in the field of AUV positioning. Unlike LBL, SBL and USBL, the multi-AUV CL method does not need to place the underwater acoustic array and surface ship in advance, which can effectively reduce the hardware complexity of the AUV localization system. The location accuracy and robustness of the system can be improved by using information sharing to estimate the position of the system, and the cooperation ability of the AUV system can be enhanced. This method is especially suitable for AUV to carry out various tasks in a large area.
In practical CL of AUVs, the measurement outliers are often mixed with the measurement information. For example, when the DVL has a water lock phenomenon, it will cause thick-tail Non-Gaussian measurement noise. In addition, the multi-acoustic propagation path between the sound source and the receiver can reflect the sound wave, causing the speed of sound to change with depth, and the reflection of the water surface and the seabed can also cause a wild value to the measurement information. In order to better deal with the measurement outliers occurring in the multiple AUVs based cooperative localization, filtering methods based on maximum correntropy criterion(MCC) such as maximum correntropy divided difference filter and maximum correntropy unscented particle filter have been proposed, and the effectiveness of the algorithms has been verified by experiments. However, they all chose different kernel bandwidths in advance in the experiment. Different error distributions and different application sites will affect the choice of optimal kernel bandwidth. The optimal value of the kernel bandwidth cannot be determined only by experience, and the unsuitable kernel bandwidth can directly affect the filtering effect. In the practical application of cooperative localization, there is no way to determine the appropriate kernel bandwidth in advance, which will greatly reduce the practical significance of MCC.
In this paper, an improved maximum correntropy cubature kalman filter(IMCCKF) is proposed to address the measurement outliers in cooperative localization of autonomous underwater vehicles (AUVs). The estimated effect of the maximum correntropy cubature kalman filter(MCCKF) algorithm is affected by the kernel bandwidth . In practical cooperative localization of AUVs, the selection value of cannot be determined only by experience, which will greatly reduce the practical application value of the MCCKF algorithm. The adaptive factor is constructed by comparing the trace size of innovation matrix and the trace size of quantity prediction error Matrix, and the kernel bandwidth in the MCCKF is corrected online by the adaptive factor. Finally, the efficacy of the proposed method is verified by the data of the lake test. The experimental results show that the proposed IMCCKF algorithm has the ability to adjust the value in real time and quickly obtain the optimal value of the kernel bandwidth, and the IMCCKF algorithm can effectively improve the positioning performance of CL system with measurement outliers.

Previous Abstract Return to Session C6 Next Abstract