Research on improved genetic algorithm for low-dimensional and multimodal function optimization

In order to overcome the disadvantages that simple genetic algorithm had in low-dimensional and multimodal function optimization, an improved algorithm was proposed. The new algorithm divided the search interval into different areas by method of orthogonal decomposition, produced initial populations that had different individual distribution density by using interval algorithm, and gave each individual an attribute that marked the area which it was in and guaranteed there was more than one individual in each area during the operating process of the algorithm. Particle swarm algorithm was introduced to the mutation operation of genetic algorithm to overcome the shortage of genetic algorithm, that was, the poor search ability in local areas, especially those that had low individual distribution density. When applied to six-hump camel back function and Branin RCOS function optimization, the simulation results show that, under the same experiment condition, in contrast to other improved algorithms, the new one improves the probability of searching to all extreme points, and has stronger ability of low-dimensional multi-modal function optimization, on the premise that the convergence precision is not affected.