Research On Collaborative State Estimation Of Unmanned Cluster With Model Risks In Underwater Complex Scenes

With the advantages of wide sensing and detection range,strong survivability,and high operational effectiveness,underwater unmanned vehicle(UUV)clusters are becoming an irreplaceable technical means for marine resource exploitation and military applications.Accurate position estimation is a prerequisite for achieving intelligent cooperation of UUV clusters,and it is also a technical bottleneck that restricts UUV clusters to engineering applications,which has become one of the hot spots of domestic and international research.However,in the complex underwater scenarios,constrained by the uncertainty of the state model,non-smooth observation noise,and weak connection of information interaction,the existing methods are difficult to meet the demand for high accuracy and strong robustness state estimation of UUV clusters.Most of the current state estimation methods are based on accurate models or Gaussian distribution assumptions,which can only handle the effects of a single factor of state model uncertainty or observation noise non-smoothness,and their performance will significantly deteriorate or even diverge when facing the model risk problem of both.In addition,research on distributed model risk state estimation methods for underwater information interaction with weak connectivity conditions is just beginning,especially in China where there is still a gap in research in this area.Therefore,this dissertation conducts three-pronged research on the significant application needs of UUV clusters to address the UUV cluster cooperative state estimation challenges for complex underwater scenarios.The main research contents and innovations of this dissertation are as follows:1.For the problem of low accuracy of navigator localization due to underwater nonGaussian noise interference,this dissertation proposes a numerically stable minimum error entropy Kalman filter state estimation method.The ocean background noise and hydroacoustic multipath effect cause the underwater unmanned cluster cooperative state estimation state space model observation noise non-Gaussian.The traditional state estimation methods are very sensitive to the outliers of model noise non-Gaussian,which can lead to great errors in the state estimation of the UUV and cannot meet the requirements of high-precision navigation and positioning.In this dissertation,a new cost function is constructed based on the error entropy,and the information entropy of the state estimation error is similarly measured by using the Minimum Error Entropy(MEE)criterion,and the outliers caused by the non-Gaussian of the model noise are suppressed by Gaussian kernel function and Parzen’s window.Moreover,the matrix singularity problem in the fixed-point iteration process is improved by the positive definite incremental construction method to ensure the convergence of the proposed algorithm.In addition,this dissertation also extends the insensitivity of the MEE criterion to the nonGaussian outliers of model noise to the distributed state estimation method,which is combined with the distributed fusion rule and the distributed diffusion rule,respectively,to obtain a high-precision cluster distributed fusion state estimation method under different communication conditions.Compared with traditional methods,the proposed method is significantly better than other algorithms in suppressing the non-Gaussian model noise and is more robust.2.The Kalman filtering algorithm framework is proposed to address the problem of unreliable estimation of vessel attitude due to the uncertainty of the vessel state model.Underwater complex scenarios of current disturbance and flexible maneuvering of the vehicle can cause uncertainty of the vehicle state model,which in turn leads to unreliable state estimation.At present,most of the model uncertainty state estimation methods are based on the assumption of Gaussian distribution,and when facing the non-Gaussian case of model noise,the existing methods cannot tolerate the interference of non-Gaussian noise outliers,which makes the estimation results seriously deviated and unreliable.This dissertation combines the information-theoretic learning optimization criterion with the game state estimation method,uses the Wasserstein ambiguity set as a measure of tolerating model uncertainty,and constructs a cost function for the nonconvex game problem based on the Maximum Correntropy Criterion(MCC)and the MEE,which not only suppresses the outliers caused by the model noise non-Gaussian but also tolerates the model in a certain radius of the Wasserstein ambiguity set content uncertainty.The numerical simulation samples and the experimental results of the UUV prove the proposed method’s effectiveness.The Kalman filtering algorithm of the model risk game proposed in this paper solves the model risk state estimation problem of model uncertainty of UUV clusters and model noise non-Gaussian at the same time to some extent.3.For the problem of weak robustness of unmanned cluster cooperative state estimation caused by weak connectivity of hydroacoustic communication,this paper proposes a distributed model risk state estimation method under bandwidth constraint.When UUV clusters interact with data in a hydroacoustic communication network,the short distance and bandwidth constraint of hydroacoustic communication due to the slow propagation speed and fast energy decay of sound in water severely limit the application of traditional cluster cooperative state estimation methods in UUV clusters.In this dissertation,the model risk state estimation results of local navigators are selected and compressed according to the coding-based method to meet the requirements of bandwidth-constrained hydroacoustic communication,and then,the incomplete information of neighboring navigators is complemented at the fusion node,and the distributed information fusion of the complemented neighboring navigators is performed with distributed diffusion rules to finally obtain the distributed model risk state estimation results under the bandwidth constraint.The method proposed in this paper solves the problem of distributed model risk state estimation under the bandwidth constraint of the UUV cluster to a certain extent.In summary,based on the practical engineering problems of UUV cluster collaboration in underwater complex scenes,this dissertation condenses the problem of model risk state estimation and communication bandwidth constraint,proposes a model risk game Kalman filter algorithm framework,gives a distributed model risk state estimation method under bandwidth constraint.It provides new possibilities and has important theoretical value and practical significance for UUV cluster intelligence research.