| The underwater vehicle is widely applied in both military and civilian due to its invisibility and mobility.With the improving requirements for navigation positioning accuracy,single underwater vehicle navigation system is difficult to meet the actual demand.Therefore,in this paper,the particle filter and its improved algorithm are researched for the underwater vehicle integrated navigation system.The major research works are as follows:1.The classic nonlinear approximation filtering method is analyzed based the Bayesian filtering estimation theory,and the particle filter algorithm theory is major researched subject.The optimization processing is against the disadvantage of the particle filter algorithm,the improved algorithm is simulated to verify the reasonableness and correctness.The research on the improved method is achieved as follows:(1)The importance density function of particle filter is designed by the state estimation mean and variance which are obtained by Cubature Kalman filter.The particles' degradation is reduced and the algorithm estimated accuracy is increased effectively by this method.(2)The nonlinear state variable dimension of the system is reduced by marginalization technique,then the calculated amount of the particle filter algorithm is reduced and the estimated accuracy is increased.(3)The regularization method is used in the resample process of particle filter to effectively restrain the depletion phenomenon of particle samples.2.Since the stationary base alignment could not be completed when the underwater vehicle is affected by the external interference,the ancillary information obtained by Doppler log is applied to the initial alignment on underwater vehicle moving base.Firstly,the error model of inertial alignment is established with strapdown inertial navigation error model and the differential equation of Euler platform angle error in the case of that the azimuth misalignment is large.Secondly,in order to overcome the effect of the observation gross errors and the system noise on state estimation,robust adaptive CPF algorithm which is based on the robust estimation and adaptive filter theories is proposed.Finally,according to the simulation results,both the algorithm stability and the high estimation accuracy are verified.3.In view of the problem that GPS navigation equipment cannot be used underwater,the integrated navigation method which is based on the inertial navigation system and long-base line(LBL)positioning system with buoys network is researched in this paper.Firstly,the working process of underwater GPS system is introduced and then the mathematical model of SINS/LBL integrated navigation system is established.Secondly,based on the characteristics of SINS/LBL integrated navigation system model,the marginalization technology is applied for separating the linear and nonlinear parts of the state variables.The state of the nonlinear part is estimated by the adaptive particle filter to further reduce the number of samples.Finally,the effectiveness of RB-APF algorithm in the SINS/LBL integrated navigation system is verified by the simulation analysis results.4.Gravity anomaly aided inertial navigation system is researched.Firstly,the Earth's gravity field model and numerical solution of gravity potential coefficients are investigated.The global gravity field model potential coefficients are improved by the local gravity field data to obtain the gravity field model which is suitable for local area.Secondly,because the measurement equation of the inertial navigation or the gravity anomaly inertial navigation system can not be expressed accurately,the Markov chain Monte Carlo method and regularization method are applied in the particle filter resample process for increasing the particle variability.Finally,according to the simulation analysis,the availability of MCMC-RPF algorithm for inertial navigation/gravity anomaly inertial navigation system is verified.5?The multi-underwater sensors integrated navigation system and the information fusion technology are further researched.Firstly,working principle and error model of underwater navigation sensors are studied for establishing the mathematical model of integrated navigation system.Secondly,because of the non-linear course in the information fusion process,the mixed federal filter is proposed based on the traditional federal filter.And then the improved Gaussian particle filter is applied in non-linear subsystem.Finally,the simulation analysis results show that mixed federal filter has a wider application prospect in multi-sensor information fusion technology. |
PDF Full Text Request