Research On Technology Of AUV Integrated Navigation System Based On Multivariate Information Fusion

As the exploration and development of ocean progresses, the demand ofunderwater vehicle with autonomous navigation capabilities is growing. In a complexenvironment of deep sea, the inertial navigation using separate components can’tovercome the system error accumulation problem, and cannot satisfy the requirementof high precision autonomous navigation, so using multiple information fusionprovided by sensors and optimal filtering algorithm become the main trend ofresearch.This article research integrated navigation algorithm of AUV. In this paper, wedesign an AUV integrated navigation system based on multiple information fusion. Inthis system we use AHRS, gps and DVL as navigation parameters input sensors tocalculating and fixing navigation system. In order to overcome the error accumulationproblem, this paper designs the "surface correction-underwater silent hunter" mode.The strapdown inertial navigation system/GPS integrated navigation system is used insurface correction mode. In this mode we use real time data from GPS to correctspeed and position information. The strapdown inertial navigation system/DVLintegrated navigation system is used in underwater silent hunter mode. In this modewe use the data from DVL and navigation system to calculate navigation parameters.In order to keep system safe, steady and high-precision, we research and realize twokinds navigation system algorithm respectively,which are combined navigation ofAUV based on H∞robust expansion algorithm and combined navigation of AUVbased on the extended kalman filtering algorithm optimized by BP neural networkalgorithm. According to the actual operation, we analyse their advantages anddisadvantages respectively.EKF is the optimizing algorithm of Kalman filter for non-linear environment. Wedesign state equation and observation equation of EKF by strapdown inertialnavigation system and DVL error equations, which can calculate the various errors. Because it is dicky that the real running environment of AUV integrated navigationsystem, the precise mathematical model of system and the transcendent characteristicof the yawp are difficult to get, and cause EKF to get divergence. Its advantage ishigher accuracy and real-time, but to be worse in reliability.In order to solve the poor stability of extended kalman filtering and, at the sametime, guarantee the accuracy of the system,we provide the combined navigationalgorithm base on H∞robust expansion algorithm and the combined navigationalgorithm base on extended kalman filtering algorithm optimized by BP neuralnetwork algorithm.H∞robust extended filter is a suboptimal estimation algorithm based onrobust control theory that it’s tenable supposition condition is external noise energylimited, and it does not require specific mathematical statistical model of noise. Andthe filter regards noise and uncertain input as limited energy signals. So the filter isparticularly suitable for the underwater situations where the external noise is unknown.What’s more, in initial alignment of SINS, the actual results are superior to what byKalman filter.The neural network can parallel processing ability, self-learning ability anddistributed storage ability, and also has very strong fault tolerance and robustness. Theneural network can ignore modeling and feature extraction process of patternrecognition method, so can reduce error caused by inaccurate selection of model orfeature. The BP neural network is a multilayer feedforward networks trained by errorreverse propagation algorithm. This paper combine extended kalman filteringalgorithm with BP neural network algorithm, which improve the stability of thesystem and filtering precision effectively, and applied to AUV integrated navigationsystem. The experimental results show that this method improves the reliability of thesystem and ensure that real-time accurate of the AUV navigation systemrequirements.