Research On Multi-source Fusion Positioning Technology With Inertial Base For AUV Based On Factor Graph

Autonomous Underwater Vehicle(AUV)is an important platform for human exploration and development of ocean resources,and the high-precision navigation and positioning is the technology support for AUV to perform underwater operations safely and reliably.As the core of AUV underwater navigation system,the multi-source fusion positioning technology is an effective guarantee to improve the accuracy and reliability of AUV and multi-AUV cluster navigation positioning.In this dissertation,the AUV based on Strapdown Inertial Navigation System(SINS)is taken as the research object,the factor graph is used as the information fusion tool,and all available information sources are used for fast fusion and reconfiguration of multiple navigation sensors,to realize the multi-source fusion positioning,ensure and improve the navigation and positioning accuracy.Considering the different working environment and mission requirements in underwater,the plug-and-play and data smoothing functions of the factor graph are applied to single AUV integrated navigation system and multi-AUV parallel cooperative navigation system.The main research work and innovative productions of this dissertation are summarized as follows:1.Aiming at the problems of the plug-and-play and dynamic changes of the availability of asynchronous heterogeneous navigation information sensors in single AUV integrated navigation system,the single AUV multi-source fusion positioning algorithm based on factor graph was proposed.Firstly,the multi-source information fusion problem of single AUV integrated navigation system is transformed into the maximum posterior probability to solve the joint probability distribution function(pdf)of multivariate random variables using the factor graph theory,and the corresponding factor graph model is constructed.Secondly,the optimal estimation of real-time navigation states can be obtained by the passing and update of the message passing algorithm in the factor graph model.Finally,the factor graph model is optimized by the equivalent Inertial Measurement Unit(IMU)factor node,which replaces consecutive IMU factor nodes between two adjacent measurements.Simulation results show that the proposed factor graph algorithm can continuously and stably output high-precision navigation solution in real time,and effectively realize the plug-and-play between the inertial navigation system and asynchronous heterogeneous sensors,and the accuracy of navigation solution is slightly higher than that of federated Kalman filtering(FKF)algorithm under the simulation condition.Semi-physical experiment results also verify that the reliability and effectiveness of the proposed scheme.2.Aiming at the high precision real-time navigation requirements of single AUV integrated navigation system,the single AUV multi-source fusion positioning algorithm based on sliding window-factor graph was proposed.Firstly,the sliding window factor graph model of single AUV integrated navigation system is constructed by the factor graph theory.Secondly,the historical factor nodes and the estimated variable nodes are selected by a fixed-width window,and the information factors are updated by window sliding and updating.Finally,the forward and backward message passing of the factor graph are performed in the sliding window,and the smoothed optimal estimation of the current navigation state is achieved by the weighted combination of the two navigation solutions.Simulation results show that compared with the sliding window federal Kalman filtering algorithm,the proposed sliding window factor graph algorithm can improve the real-time navigation estimation accuracy of single AUV integrated navigation effectively.Semi-physical experiment results verify that the reliability and effectiveness of the proposed scheme.3.Aiming at the cooperative navigation and localization problem of multi-AUV cluster in unknown underwater environment by using asynchronous heterogeneous self-observation information and relative observations information,the multi-AUV multi-source fusion positioning algorithm based on factor graph was proposed.Firstly,the mathematical model of the multi-AUV parallel cooperative navigation system is established.Secondly,the factor graph model of the multi-AUV parallel cooperative navigation system is constructed by the factor graph theory.Then,the local navigation information of each AUV and the relative observation information between AUVs are split into each AUV platform node,which only updates parameters related to its own state.Finally,the optimal estimation of real-time navigation states can be obtained by the passing and update of the message passing algorithm in the factor graph model of the multi-AUV parallel cooperative navigation system.Simulation results show that the proposed algorithm can not only use the relative observation information between AUVs to delay the divergence of positioning error,but also fuse the local navigation information of each AUV effectively.Semi-physical experiment results verify that the reliability and effectiveness of the proposed scheme.4.Aiming at the requirements of offline high-precision navigation and positioning in the mapping of ocean floor,survey and other fields for AUV underwater navigation systems,two post-processing algorithms for AUV underwater navigation are proposed,which are the two-filter smoothing algorithm and the Rauch-Tung-Striebel(RTS)smoothing algorithm based on the factor graph.Firstly,in the framework of factor graph model,the factor graph is combined with the two-filter smoothing algorithm and the RTS smoothing algorithm respectively.Secondly,the message passing algorithm is used to pass and update messages in the factor graph model of multi-AUV parallel cooperative navigation system.Finally,all measurement information within the fixed interval is fully used to estimate the navigation state at each moment.Simulation results show that the navigation accuracy of the two-filter smoothing algorithm based on the factor graph is similar to that of the RTS smoothing algorithm based on the factor graph,and compared with the federated Kalman filtering smoothing algorithm,the navigation accuracy of the two algorithms is slightly improved and can play a good smoothing effect on the overall navigation solution results;compared with the real-time factor graph filtering algorithm,the navigation accuracy of the two proposed algorithms is significantly improved,especially when in the case of signals loss or sensors fault.Therefore,both of them are effective data post-processing methods.Semi-physical experiment results verify that the reliability and effectiveness of our two data post-processing schemes.