Design method based on worst case genetic algorithm

Abstract

The invention discloses a design method based on a worst case genetic algorithm, and relates to the field of digital filter design. According to the invention, the genetic algorithm is improved, and the optimization method under the worst condition is obtained. Firstly, the method improves a screening mode of an optimal individual in an existing genetic algorithm during solving, and obtains a method for evaluating the optimal individual under the worst condition for a single time. Secondly, the disadvantage that a gradient information-based algorithm is easy to fall into local optimum and the problem that the solution without derivative or information shortage cannot be handled are effectively avoided by means of crossover variation and the like of a genetic algorithm. The algorithm is low in calculation cost in digital filter design and wide in application range, and a better method is provided for optimization under more general conditions.

Description

technical field [0001] The invention relates to the field of genetic algorithms, in particular to a design method based on worst-case genetic algorithms. Background technique [0002] The digital filter digitally filters out the undesired signal in the signal transmission process and retains the desired signal part, so as to realize the purpose of information extraction and utilization in digital signal processing. Digital filters are the basic operating units of digital signal processing systems, and have been widely used in signal processing in industries and fields such as digital audio-visual equipment, aerospace, and medical systems. However, the current digital filter design does not consider the frequency response approximation error in the transition band, resulting in the filter amplitude response often has obvious overshoot in the transition band, and the group delay error at the edge of the passband is also large. Therefore, the design for minimizing the maximum ...

AI Technical Summary

Problems solved by technology

However, this model cannot fully describe the real situation, so the model under continuous variables is widely accepted

As for the optimization model under the completely continuous worst case, there are already surrogate model-based optimization methods proposed in articles such as "Anew expected-improvement algorithm for continuous minimax optimization", but due to the large calculation consumption of this method, it is practical Application is still difficult

For simpler worst-case optimization, "Evolutionary Algorithms for MinimaxProblems in Robust Design" proposes to use some evolutionary algorithms for optimization. The calculation method is simpler, but it still cannot solve some complex asymmetric problems.

For complex asymmetric problems, "Necessary conditions formin-max problems and algorithms by a relaxation procedure" proposes a general worst-case optimization framework, but it is limited to the optimization method based on gradient information, so it cannot be solved Problems with no derivative information and extremely high degree of nonlinearity

[0005] The above method can solve some special problems, but still lacks a more general and relatively low-calculation worst-case optimization method

Images