Modification Research On The Exploitation Capability Of The Backtracking Search Optimization Algorithm Based On Natural Inspiration

Optimization is an important branch in the fields of Applied Mathematics and Computational Science.Meta-heuristic algorithms have become one of the most popular research directions in the field of optimization.With the continuous development of computer technology,more and more novel and efficient meta-heuristic algorithms have been widely applied in many engineering applications.Backtracking search optimization algorithm(BSA)is an emerging meta-heuristic algorithm based on the population.BSA possesses a unique memory function,which enables it to utilize both current information and historical information in every evolution process.In addition,BSA also has a powerful random mutation operator and a double-crossover mechanism invoked by probability.These advantages of BSA give it a powerful global exploration capability.However,there are still some shortcomings of BSA.On the one hand,too large amplitude range of BSA's mutation control parameter could easily weaken the local exploitation capability and then affect the convergence speed of BSA.On the other hand,BSA will be easily to fall into the local optimum,when its historical information and the current information tend to be the same.Based on these considerations,two modified BSA based on natural inspirations are proposed to improve the optimization performance of the algorithm in this paper,after the improvement researches of BSA and its applications are comprehensively reviewed.The proposed modified algorithms are then applied to solve the real-life constrained engineering optimization problems.The main research works of this paper are as follows.(1)The design principle and algorithm framework of BSA are introduced,the advantages and disadvantages of the algorithm are analyzed,and the improvement researches and applications of BSA are systematically reviewed.This lays foundation for the following modification researches in this paper.To test the performance of BSA,a non-isometric point segmentation numerical integration method based on BSA is proposed to solve several examples of complex numerical integration problems.The results of these numerical examples indicated that the performance of BSA is more competitive than the similar algorithms.(2)The mutation control parameter(F)of BSA has a considerable fluctuation range and that will affect convergence speed of the algorithm.This paper proposed a modified BSA inspired by simulated annealing(BSAISA)to overcome this deficiency of BSA.In the BSAISA,an adaptive F is redesigned by through learning the characteristic of the acceptance probability in simulated annealing.The new F could be adaptively decreased as the number of iterations increases.The modified strategy can provide an effective trade-off between the global exploration in the early iterations and the local exploitation in the later iterations for the algorithm.The high performance of the modified algorithm is verified on two constrained optimization test sets.(3)In order to further improve the performance of BSA,on the basis of the BSAISA,another natural inspired improvement strategy is introduced,namely species evolution rule.Thus,a novel modified BSA inspired by simulated annealing and species evolution rule(SSBSA)is proposed.The new modified strategy uses the specified retain mechanism from the species evolution rule for reference.With the guidance of fitness feedback information,the historical population(oldP)and the parameter(F)of the previous generation are retained to the next iteration.This unique retain mechanism could improve the local exploitation capability of BSA and effectively to make BSA avoid falling into local optimum.SSBSA has been carried out on simulation experiments of on some constrained engineering optimization problems.The experimental results verify that the performance of SSBSA is better than those of BSAISA and BSA.Moreover,the experimental results of SSBSA and other similar algorithms are compared,it verifies that the SSBSA is more competitive than other algorithms in terms of convergence speed.