一种针对全海深载人潜水器的异步融合组合导航算法

An Integrated Navigation Algorithm with Asynchronous Fusion for Full-Ocean-Depth Human Occupied Vehicle

全海深载人潜水器(HOV)组合导航中会产生异步融合现象,传统的组合导航算法在处理时会产生较大的误差.针对这一问题,提出了一种基于机器学习和无迹卡尔曼滤波(UKF)的异步融合组合导航算法.首先建立了针对超短基线(USBL)声学定位系统预测的机器学习模型,通过USBL声学定位系统的观测数据集来训练该模型,并用得到的模型来预测更新间隔内的数据.最后使用UKF将已更新的数据集进行融合.仿真结果表明,相比传统的组合导航算法,本文的异步融合组合导航算法可以将USBL声学定位系统数据异步问题所引起的误差降低17%,有效提高了组合导航系统的精度.

The asynchronous fusion maybe happen in the integrated navigation of full-ocean-depth human occupied vehicle(HOV),and large error can be caused if the traditional integrated navigation algorithms are used.To solve this problem,an integrated navigation algorithm with asynchronous fusion is proposed based on machine learning(ML)and unscented Kalman filter(UKF).At first,an ML model is established for prediction of the ultra-short baseline(USBL)acoustic positioning system.Then,the model is trained by the observation dataset of USBL acoustic positioning system,and the data in the intervals between updates are predicted by the model.Finally,the updated dataset is fused by using UKF.The results of simulation experiments manifest that compared with the traditional integrated navigation algorithms,the error caused by asynchronous data from USBL acoustic positioning system can be reduced by 17%,by the proposed integrated navigation algorithm with asynchronous fusion,and the accuracy of the whole integrated navigation system is effectively improved.