Research Of Bat Algorithm Based On Doppler Compensation And Mutation Selection

In order to overcome the problems of bat algorithm(BA)such as falling into local optimal value and unstable convergence,this paper proposes a bat algorithm based on Doppler compensation and mutation selection(BADM).First,in the global exploration stage,Doppler frequency compensation strategy and Doppler velocity compensation strategy are proposed.The Doppler frequency compensation strategy increases or decreases the frequency appropriately according to the distance between the individual and the optimal individual,prevents random change and makes the algorithm more consistent with the physiological mechanism of the bats,so as to provide guidance for adjusting the velocity.The Doppler velocity compensation strategy uses the average population velocity to guide the direction of individual velocity,adjusts the displacement increment and corrects the value of the velocity with the compensated frequency to generate new individuals close to the optimal individual,so as to avoid the blindness of global exploration.Second,in the local exploitation stage,a mutation selection strategy is proposed.It uses Cauchy mutation to generate a larger step to expand the search area and then uses Gaussian mutation to generate a smaller step to search the optimal area finely so as to promote the individual to adjust the step size and break through the constraint of local optimal value for increasing the flexibility of mutation.Third,a linear decreasing strategy for the pulse rate is proposed to balance the global exploration and local exploitation.Fourth,a linear decreasing strategy for loudness is proposed to increase the probability of retaining new individuals in the global exploration and local exploitation,so as to increase the rate of population evolution.After analyzing the convergence and computational complexity of BADM theoretically,a number of comparison experiments between BA with each strategy and original BA were set up for verifying the effectiveness of the improved strategies of BADM,which proved the feasibility of each strategy to improve the performance of BA.In order to verify the optimization performance of BADM,an experiment was set to evaluate the value range of pulse rate and loudness,and then 12 standard test functions were used to carry out simulation comparison experiments under low and high dimension.The results show that BADM has higher convergence accuracy and lower convergence standard deviation than BA.There are 35 figures,14 tables and 71 references in this paper.