Research On Array Pattern Synthesis Based On Intelligent Optimization Algorithms

The array pattern synthesis is a research hotspot in the field of array signal processing.In order to obtain the expected array pattern,the variable parameters in the uniform and phased arrays are optimized for the low sibelobe level(SLL),accurately controlled null,and high directivity.These methods are widely used in many systems with large scale array,such as communication systems,radar systems,phased array systems and so on.The evolutionary algorithms have made great achievements in array pattern synthesis,but there are still many problems on the convergence performance and multi-objective optimization for these methods,especially for the application in the array pattern synthesis.Based on the intelligent optimization algorithms,the main research contents are to improve the convergence and the diversity,especially for the application in the array pattern synthesis.The research in this thesis can be shown as: the research of the particle swarm optimization with feedback value and nonlinear dynamic weight;the research of multi-objective optimization methods with interpolation or fitting and the diversity of optimal solutions;the research of the method of digital position shift with virtual sparse elements for low SLL in beam scanning.The main contributions of this thesis are as follows:(1)For the problem of poor convergence in multi-objective array pattern synthesis by particle swarm optimization(PSO)method,a feedback PSO with nonlinear dynamic weight is proposed to search in a large initial space and converge fast in the local space to a refined solution.The dynamic weight is determined by a subtriplicate function with feedback taken from the fitness value of the previous best solution.The dynamic weight is adjusted in real time with the evolution process.Numerical examples illustrate the high performance of the convergence in the array pattern synthesis by the proposed method.The optimized pattern has the accurate nulls at every specific direction and demonstrates inhibiting ability for the SLL.(2)For the performance degradation of SLL in beam scanning by the sparse array,a digital position shift method(DPSM)with virtual sparse elements is proposed to suppress the SLL when considering all possible steering directions for beam scanning.In the condition of fixed physical positions of the array elements,the virtual elements can be calculated by exponential phase synthesis after optimizing position offset factor of each element.The position shift range can be constrained by analyzing the correlation of virtual array receiving signals.Numerical examples show that the proposed method can achieve the pattern with a lower SLL in different steering directions,compared with the methods used for sparse array pattern synthesis.(3)To improve the uniform diversity for the solutions of multi-objective optimization methods and apply multi-objective optimization methods to the problems of multi-objective array pattern synthesis,the multi-objective genetic algorithm based on fitting or interpolation(MOGA/F and MOGA/I)are proposed.The expected reference points are uniformly distributed in the objective space,which is calculated by applying a fitting function or interpolation method for the objective values in the same nondomination rank.The error matrix can be constructed by the Euclidean distances between the population and the reference points.Only one individual corresponding to each reference point will be copied to the new population,according to the proposed error-comparison operator.In this thesis,MOGA/F and MOGA/I are compared with the traditional multi-objective optimization methods by optimizing the mathematical problems.The numerical examples illustrate that the proposed methods can maintain the performance of uniform diversity without destroying the convergence.(4)For the problem of poor diversity of the solutions in multi-objective array pattern synthesis,an improved nondominated sorting multi-objective genetic algorithm is proposed with three modifications,which are dynamic nondomination strategy,scope-constrained strategy,and front uniformly distributed strategy.For a large search space in the initial process,dynamic nondomination factor is considered in the rank operator.A manageable number of Pareto solutions in the constrained scope can be used to take the place of the entire Pareto front to reduce the size of population and the computation complexity.In the selecting operator,the solutions closer to the uniformly distributed points on the current front will be chosen and added to the new population,which can improve the diversity of the final solutions.The proposed methods and two efficient multi-objective optimization methods are used for the optimization of mathematical problems and array pattern synthesis,combined with the pattern of real antenna in engineering application as well.The numerical examples illustrate that the proposed method has the better performance in computation complexity,convergence and the diversity of the final fitness function values.