A Parallel fusion method for AUV cooperative localization based on weighted information gain

In order to improve the global localization accuracy and the real-time localization information in the cooperative localization of the autonomous underwater vehicle (AUV) system under clutter interference, a local information fusion algorithm based on information gain is proposed.

The gross error of the observed value is improved by the threshold weighting method, and then the local information is filtered to optimize the observed value so that it is closer to the true value. In this paper, the reliability of each piece of measurement information is investigated from the information entropy theory, and multiple sets of local filtering information of multiple master underwater vehicles are optimized. Taking the information gain as the weight, multiple sets of filtering results are fused to generate the unique localization information of the tested slave AUV. Furthermore, due to the communication delay in sonar detection and underwater acoustic signals between master and slave AUVs, filtering data asynchrony may occur in local information filtering and the information gain fusion algorithm. In view of this, a parallel structure of local information filtering and information gain weighting is proposed, which utilizes a real-time update mechanism to ensure that the input values of the information weighting algorithm are the latest output values of local filtering.

The simulation results show that compared with multi-source local filtering information, the proposed fusion method can effectively reduce the absolute error of local filtering, improve the localization accuracy, and optimize the local filtering.

The proposed fusion method can effectively realize the cooperative localization of the multi-AUV system.