Research On Cooperative Localization Algorithm For Multiple Autonomous Underwater Vehicles

"Cooperative localization of Multiple Autonomous Underwater Vehicles(AUVs)”is a new underwater positioning technology which developed with the collaborative application of Multiple AUVs.It not only can achieve high precision underwater navigation,but also low-cost,simple implementation,good reliability and without restricted operating area.It has become a new hotspot of underwater navigation.Due to the complex underwater acoustic communications environment,and unique characteristics of underwater acoustic communication technology,different from traditional land(or air)collaborative localization technology,the AUVs cooperative localization has its unique characteristics and difficulties.To improve the performance of AUVs cooperative localization,this work mainy focus on the problems of AUVs cooperative localization based on acoustic range measurements.The main works are as follows:1?The principle of AUVs cooperative localization based on acoustic range measurements is introduced.In view of weak observability due to limited observed information,observability analysis of cooperative localization system is presented based on linearized system theory and nonlinear Lie derivative theory respectively.To further clarify the relationship of observability metric and relative motion of AUVs,quantitative analysis based on the matrix condition number is presented,and then the program of cooperative system can be designed based on which.Finally,the program of cooperative system with two leaders is designed based on the theoretical basis of the observation.2?From Bayesian estimation viepoint,two kinds of nonlinear filtering algorithm are introduced,including Extended Kalman Filter(EKF)and Divided Difference Filter(DDF)based upon the principle of linearzing the nonlinear system models and Unscented Kalman Filter(UKF)based upon the propagation of a cluster points.By comparing the cooperative performance for these filtering algorithms through simulation,theoretical foundation is provided for the algorithm choice in subsequent cooperative localization use.To improve the cooperative performance in conditions of low frequency update rate and weak observability,and improve the convergence speed in large initial error.The iterated filtering algorithm is proposed to collaborative localization system.The corresponding iterated filter algorithm is designed,and then by using the Levenberg-Marquardt(L-M)method to the iterated filter algorithm,the stability of the iterative filtering algorithm is improved further.Finally,the simulation is carried out and shows that the method is efficient.3.To improve the performance in condition of outliers,particularly to avoid serious cumulative amplification of the estimation error for the iterated filtering algorithm,a robust Huber-based iterated filtering algorithm is proposed for the cooperative position estimate.The principle of the Huber robust estimate algorithm is introduced first in detail and then a modified robust algorithm is presented by considering the leverage points in design matrix.By identifying and weighting the leverage points,the stability of the robust estimate algorithm is improved.Based on the Huber robust estimation theory,the Huber-based EKF and DDF filtering algorithm are designed,and then the robust iterated filtering algorithm is proposed for underwater Cooperative localization.Finally,simulation experiments are carried out in condition of single outlier measurement noise and heavy-tailed noise respectively,simulation results show that the robust iterated method is validity.4.The noise often exhibit uncertainty or variability in actual system.To improve the noise adaptive ability for nonlinear filtering algorithm and improve the estimated position accuracy,an adaptive filtering algorithm combined with Myers-Taple method is proposed on the basis of robust filtering algorithm.First,the Myers-Tapley noise adaptive estimation algorithm is introduced,and then a modified Myers-Tapley method with weighted residuals is proposed to improve the performance in non-Gaussian condition.In order to smooth the estimate histories,take into account the priori statistical characteristics of the system noise at the initial time,a fading memory method is used to ensure the stability of noise adaptive estimation results.Finally,simulation experiments were carried out in condition of Gaussian and non-Gaussian noise respectively,simulation results show that the robust iterated method is validity.5.To verify the performance of collaborative navigation algorithms in real condition,lake trials based on acoustic communication equipment were conducted using surface boats.The program of Cooperative localization trial based on two leaders was introduced first and then offline test data were used to simulate the cooperative algorithm for single and two leaders respectively,as well as in different measurement update cycle.