Research On The ADMM-Based Algorithm For Large Scale Array Pattern Synthesis

With the development of modern electronic technology and computers,large-scale array have been widely used in modern radar,communication,sonar and other fields due to its high resolution,high gain,and high signal transmission rate.As a key technology of array signal processing,array pattern synthesis has become one of the current research hotspots.With the increase of array size,the traditional array pattern synthesis method is hard to solve the highly non-linear model and high computational complexity problem.In view of these difficulties,combining with modern optimization theory,especially largescale parallel optimization theory,this dissertation focuses on the analyzing and designing large-scale array pattern synthesis methods and efficient algorithms.The research contents of this dissertation mainly include: The conventional large-scale array pattern synthesis technology based on ideal steer vector and the robust large-scale array pattern synthesis technology.For nominal array pattern synthesis problem,this dissertation establishes minimizing sidelobe level model,maximizing directivity model and multi-objective pattern synthesis model based on the ideal array steer vector,which are proposed for designing array pattern with low side lobe,narrow beam.To directly controal the main lobe width,an improved general-purpose model is proposed based on weighted steer vector.For large-scale array scenarios,most of the existing pattern synthesis methods usually face difficulties in high computational complexity,which leads to the applicable scenarios are greatly reduced.Therefore,this dissertation designs a low complexity,parallelizable optimization algorithm based on the alternating direction method of multiplier(ADMM)to obtain the global optimal of array pattern synthesis problem.In the simulation,the large-scale planar array antenna system is empolyed,the results verify the validity of these models by analyzing different parameters in the models.For the robust pattern synthesis problem,aiming at the difficulty of measuring steering vector error.In this dissertation,the amplitude perturbation and phase disturbance of steering vector are used as error source,and a new steering vector with multiplicative error model is employed,of which the error are measurable.However,the nonlinear multiplicative error model is not conducive to solve pattern synthesis problem.so this dissertation transforms the multiplicative error model is into additive noise model by relaxation technique.Based on this model,a robust pattern synthesis model with minimized sidelobe level is proposed based on worst case criterion,and the non-convex model is transformed into a second-order cone programming problem by using the convex relaxation methods,which can be solved in polynomial time.Additionally,this dissertation presents a parallelizable algorithm based on the idea of ADMM,which can solve the robust large-scale array pattern sythesis model globally.The somulation results show that the robust pattern synthesis methods based on the additive noise model can greatly reduce the peak sidelobe level,and also can select antennas for large-scale array.On the other hand,the convergence results verify that the proposed ADMM algorithm is characterized by low complexity,high efficiency.