Research On Improvement Of Bat Optimization Algorithm

Swarm intelligence algorithm has become a hot research scholar in recent year and it has broken the limitations of traditional numerical calculation method.The bat algorithm derived from bat predatory behavior and echolocation by professor Yang and had some advantages such as simple structure,less parameters,strong commonality and so on,it has been successfully applied in assembly line scheduling,path planning,network node partition,image compression and other practical problems.But it had some deficiencies like other swarm intelligence algorithm,such as low accuracy in searching for the optimal solution area,easily trapped in local optimum,slow convergence speed and so on.The simulation indicated that the speed of search optimal solution area is very slow in the early stage.It required hundreds of iterations to search the whole space.With the increase of the number of iterations,the population was clustered toward a position,and the population aggregation occurred.Then Increased the number of iterations,found that almost all the individual position never change.This was because the position formula was constrained by the current moving speed,and the current moving speed decreased with the number of iterations.After fully understanding the reasons of the slow convergence speed and low accuracy,the improvement on the basic bat algorithm are as follows:(1)This paper added inertia weight in the speed formula which obeyed the uniform distribution and beta distribution,thus accelerated the convergence speed.In addition,this paper designed a speed correction factor and used it to constraint the step of bat dynamically,which provided the algorithm with effective stability and adaptability.An adaptive bat algorithm with dynamically adjusting inertia weight(DAWBA)was proposed,Simulation results indicated that the performance of DAWBA is improved.(2)In order to solve the phenomenon of population aggregation,uniform mutation and Gaussian mutation was integrated into the whole search procedure,two mutation mechanisms allowed the algorithm to locate global optimal space quickly and improve the accuracy of the solution.Meanwhile,a mutation switch function which made all individuals have the probability of mutation operation in the whole process was designed to ensure the diversity and activity.Bat optimal algorithm combined uniform mutation with Gaussian mutation(UGBA)was proposed,Simulation results indicated that the performance of UGBA is significantly improved.