基于At-LSTM的GNSS失锁环境定位技术研究

打开文本图片集

关键词：组合导航；GNSS失锁；卡尔曼滤波；长短期记忆网络；注意力机制DOI：10.15938/j. jhust.2025.05.012中图分类号：TN967 文献标志码：A 文章编号：1007-2683（2025）05-0114-10

Abstract：ToaddresstheissueofdecreasedpositioningaccuracyinGNS-denedenvironentsfortheGlobalNavigatioSatelite System/Strapdown InertialNavigationSystem（GNSS/SIS）integratednavigationsystem，anintegratednavigationsystemaistedby Atention-basedLong Short-TermMemory（At-LSTM）network isproposedbasedontheKalmanfilter modeloftheintegrated navigationsystem.The network is trainedusing measurementdata when GNSS signalsareavailable.In GNS-deniedenvironments， thetrainednetwork isusedtoreconstructGNSSsignalsbasedonSINSmeasurements，and positioningisthenperformedthroughthe Kalmanfiltermodel.Theatentionmechanismisintroducedtodistinguishthimportanceofnetworkinputstotheoutput，tereby ensuringthestabilityoftheintegratednavigationsystem.Experimentswereconductedonamobileexperimental platfominGNSSdeniedenvironments.Theesultssowthatcomparedwithpureinertialnavigation，theAt-LST-asistedintegatednavigatiosystem reduces the east and north positioning errors by 54.8% and 18.2% ，respectively，after 3O seconds of operation，and by 81.5% and 67.3% ，respectively，after9Osecondsofoperation.Thisefectivelypreventsthedivergenceofpositioningresultsandsignificantly improves the positioning accuracy of the integrated navigation system in GNSS-denied environments.

Keywords：integrated navigation；GNSS-denied；Kalman filtering；long short-term memory；atention mechanism

0 引言

当今，智能移动机器人[1-3]和无人移动载具[4-6]获得了广泛的应用，定位技术能够给出物体在空间中的具体位置信息，是这些应用的基础。（剩余13640字）