Improvement And Application Of Multi-Verse Optimizer

S Mirjalili et al.have been inspired by the multi-verse theory of physics and proposed a new swarm intelligence optimization algorithm called multi-verse Optimizer(MVO)in 2015.This algorithm has the advantages of simple framework,few controlled parameters,self-organization and self-adaptability,but it also has the disadvantages of late convergence delay,low solution accuracy and easy to be trapped in the local optimum.Based on the shortcomings of MVO algorithm,this dissertation proposes some improvement measures to improve the performance of MVO algorithm and expand its application field.The main research work is as follows:1.For the original MVO algorithm,its linearly increasing WEP is replaced by log-increasing LWEP,and then adaptive constrict factor and Cauchy mutation operator are introduced.LWEP increases the probability of updating for universe individual.Adaptive constrict factor can effectively balance the global convergence and local convergence of MVO,and Cauchy mutation operator makes MVO escape from the local extreme value.The effectiveness of the improved algorithm is verified by function optimization and engineering examples.2.The idea of complex-valued encoding is used to improve the MVO.Each universe individual is represented by two variables,the real part and the imaginary part of complex number.The diploid encoding of complex number can enlarge the search space,enrich the population size,enhance the global search ability of MVO,and help MVO jump out of the local optimization.The improved algorithm is applied to IIR system identification,and the experimental results show that a better solution can be found.3.A chaotic Multi-Verse Optimizer based on logistic chaotic map is proposed.The randomness,ergodicity and sensitivity of initial conditions of logistic chaotic map are helpful for MVO to perform global search behavior and avoid the trouble of local optimization.The improved algorithm is used to solve the parameter tuning of PID controller in AVR system,and the comparison of experimental results shows that it has better performance.4.Differential mutation operator of Differential Evolution is inserted into MVO.The population variation mechanism can enrich the diversity of multi-verse population,improve the local search ability,and overcome the obstacle that the original MVO is prone to fall into the local optimization within a short iteration range.The improved algorithm is successfully applied to three different dimensional cases of UCAV path planning.