Multipath Estimation Based On Information Theory Learning

It is the so-called technology changing our lives.The satellite navigation system can provide full-time and multi-directional positioning service,which greatly facilitates the acquisition of spatial location information.The positioning accuracy of navigation system will directly affect user experience to a large extent.Multipath interference has become the main error source in the field of high precision positioning because of its spatiotemporal independence.In recent years,multipath suppression based on data processing has attracted much attention with flexibility and ease-of-development,which core is multipath estimation.Common multipath estimation methods based on data processing are only efficient in Gaussian noises since they only consider the second-order statistical information of estimation errors.However,the electromagnetic environment is complex in the actual environment.The non-Gaussian noise interference,such as impulse noise,often exists in the environment where the receiver is located.The performance of traditional algorithms is obviously degraded or invalid under non-Gaussian noise.Therefore,it is necessary to improve the environmental adaptability of traditional multipath estimation algorithm,which has great significance for the study of multipath estimation under non-Gaussian noise.In recent years,information theory learning(ITL)has been widely used in parameter estimation due to its strong robustness and universality.The basic idea of parameter estimation method based on information theory learning is using probability theory and mathematical statistics to study information entropy,and to measure the randomness of estimation error with various entropy indexes.Therefore,this paper mainly considers the problem of multipath estimation under non-Gaussian noise by using the information theory learning method.Aiming at the multipath estimation problem under non-gaussian noise,the following work is completed in this paper:1.A multipath estimation algorithm based on survival information potential(SIP)is proposed to solve the problem that the traditional multipath estimation algorithm based on mean square error(MSE)is not suitable for non-Gaussian noise.The algorithm uses SIP as the objective function to optimize the estimation results that can simultaneously measure the randomness and convergence of estimation error,using the stochastic information gradient method to minimizing the SIP to ensure that the estimation error is near to zero while having minimum randomness.A multipath estimation filter based on SIP is proposed,and the iterative estimation of multipath parameters is realized.Compared with the other two multipath estimation algorithms based on ITL: minimum error entropy(MEE)and centered error entropy criterion(CEEC),the proposed algorithm in this paper has the advantages of easy calculation,higher real-time convergence and faster convergence since without calculate the kernel function and the selection of the kernel width.The theoretical and simulation results verify the effectiveness of the proposed algorithm in multipath estimation under non-Gaussian noise.2.The generation and disappearance of multipath always change with the motion of the receiver in the real environment,so the parameters and number of multipath also change.However,most of the existing literature only studies static multipath.This paper studies the dynamic multipath estimation problem under non-gaussian noise.Firstly,maximum correntropy criterion(MCC)and generalized maximum correntropy criterion(GMCC)are integrated into the traditional adaptive multipath estimation(AME)algorithm,named MCC-AME and GMCC-AME,in order to realize the dynamic estimation of multipath number and multipath parameters under non-gaussian noise.Then,introducing the forgetting factor,the recursive maximum correntropy criterion(RMCC)and recursive generalized maximum correntropy criterion(RGMCC)are used as the performance index of adaptive multipath estimation for improving the estimation accuracy and reducing the time consumption,named RMCC-AME and RGMCCAME.The proposed algorithm can realize dynamic multipath estimation quickly and efficiently without assuming a fixed number of multipath.At the same time,it can effectively overcome the initial state sensitive problem of the multipath estimation algorithm based on ITL previous proposed by the research team.Finally,the performance of the four algorithms is compared from the aspects of implementation process,convergence speed and estimation accuracy.The results show that RGMCC-AME algorithm has the best dynamic multipath estimation performance.The main thesis of this dissertation comes from the project of National Natural Science Foundation of China(No.61603267,No.61503271).The purpose of which is to provide a new research idea for multipath estimation under non-Gaussian noise,which is of great theoretical and practical significance for the development and application of high-precision navigation.