A Study On The Inversion Algorithm Of Atmospheric Duct In The Troposphere

Atmospheric duct in the troposphere may significantly affects the propagation path and range of electromagnetic waves,thereby affecting the operational performance of radar and other radio electronic systems,and restricting the tactical deployment and selection of electronic systems.How to use the information carried by electromagnetic waves propagating in the atmosphere to invert atmospheric duct is a hot issue in inverse problems research.In this paper,the systematic study of GNSS signal joint inversion atmospheric duct technology is carried out,and the adjoint mode of non-local boundary(NLBC)is derived.The main contents are as follows:1.Aiming at the problem that the fixed parameter value of the cuckoo algorithm limits the performance and convergence speed of the algorithm,a feedback adjustment parameter scheme based on evolutionary process is introduced,and a dynamic adaptive adjustment cuckoo algorithm(DACS-CO)is proposed.The comparison of different swarm intelligence algorithms in various nonlinear systems shows that the algorithm can improve the convergence speed and solution quality of the cuckoo algorithm through the dynamic adaptive adjustment and crossover mutation of the parameters,and further improve the diversity and intensiveness of the population to effectively invert the parameters of the nonlinear system.Under the condition of horizontal uniformity,the algorithm is used to carry out inversion of evaporation duct and surface duct.The numerical experiments prove that DACS-CO has fast convergence speed,strong anti-noise performance and can achieve the purpose of real-time inversion.2.The framework of combining GNSS propagation loss with signal delay information to jointly inverting atmospheric duct was constructed,and according to the deficiency of classical multi-objective optimization algorithm NSGA-?,an improved algorithm NS SAGA combined with simulated annealing was proposed.When the objective function is the same,the inversion result of the hybrid algorithm is closer to the simulated true value than NSGA-?.Moreover,under different Gaussian noise environments,the comparative analysis results show that NS SAGA has strong anti-noise ability.When the hybrid algorithm is adopted and the objective functions are the matching field processing method and the ordinary least square method,the experiment shows that the result of the matching field processing method is better than the least square method.Bartlettl,which considers the influence of antenna height,can obtain better inversion effects under different Gaussian noise levels.For the actual observation data,this paper uses Poyang Lake observation data to further test NSSAGA combined with Bartllettl.The results show that the proposed new inversion method can better describe the real refraction environment.3.GNSS signal joint inversion is a multi-objective optimization problem.Using HV,IGD and ?2 as an evaluation index,the diversity and convergence of the solution set of 7 evolutionary multi-objective algorithms are evaluated from the perspective of whether the antenna height and the transmission frequency are considered to select a multi-objective algorithm that is suitable for solving such problems.Using HV as the evaluation index,the results showed that NSGA-? performed best in the three groups,followed by NSGA-?,and MOEAD performed the worst.With HV and IGD(or ?2)as metrics sometimes,evaluation results may vary,especially for MOEAD,HypE and GrEA.When the antenna height and transmission frequency are not taken into account,KnEA fails to converge the inversion results to the true value,and the diversity of target space is weak.When considering the transmission frequency,KnEA improved the inversion effect significantly,and the target space was evenly distributed on PF.When two factors are considered simultaneously,the target space converges and the diversity of decision space is enhanced.Comparing the inversion results with the simulation results,the results show that some inversion algorithms are not necessarily close to the real value,even if they show good performance on the evaluation indicators,such as HypE.In contrast,MOEAD and NSGA-? can more accurately construct the atmospheric refraction environment under different conditions,indicating that the two algorithms are more robust.4.An adjoint inversion algorithm combining non-local boundary condition(NLBC)is proposed.The gradient of refractivity parameters to be inversed with respect to the target function is calculated by adjoint mode.Finally,the retrieved profile is obtained by the quasi-Newton gradient descent method.Compared with APM,NLBC has good absorption effect and can achieve the same purpose as Hanning window function.In the adjoint code test,as ? decrease.?1(?),?2(?)and ?3(?)are close to 1.It indicates that the tangent linear model and adjoint model code are designed correctly,and the obtained functional gradient can be used to optimize the control.During the evaporation duct retrieved,the low-frequency electromagnetic wave has the same effect in two different parametric forms of evaporation duct,while the relatively high-frequency electromagnetic wave performs poorly in the logarithmic evaporation duct.In the numerical test of anti-noise performance test,the inversion algorithm exhibits ill-posedness in high-level noise,and the inversion parameters below 10m oscillate around the true value.