Improvement Of Cuckoo Search Algorithm And Its Applications To Distribution Routing Problems

The optimization problem is a common mathematical programming problem in algorithm research.It is also one of the most active and favored research fields by researchers.In recent years,the scale of problems is more and more intricate,and the traditional methods of numerical calculation have problems of the time-consuming and solving difficulties.It cannot satisfy the needs of people in the reasonable time.Therefore,in order to find a new solution,researchers began to pay attention to the behavior patterns of natural biological groups,and propose many intelligent optimization algorithms.However,the development of these intelligent computing methods has provided new ideas for the solution of optimization problems and has been applied to various fields widely.In 2009,Yang and Deb proposed a new global search algorithm—cuckoo search algorithm by mimic the breeding behavior of cuckoos.This algorithm makes the global and local exploration effectively through the lévy flight mechanism and random walk strategy.Cuckoo search algorithm has become a research hotspot in the field of intelligent optimization algorithms because of these advantages of simple model,low parameter setting and strong universality since it is proposed.In recent years,the algorithm has been successfully applied to these fields of engineering optimization,route planning,resource scheduling and image segmentation etc.with the further research and deepening of algorithms.But this algorithm is also the same as other intelligent algorithms,and there are some problems of slow convergence rate and premature convergence in the search process of optimal solutions.Aiming at these shortcomings,this paper detailed analyzes the causes of the phenomenon by consulting a large number of related documents and experimental testing,and makes corresponding improvements to the above-mentioned phenomena in the algorithm.The improved algorithm is applied to the problem of cargo distribution route planning.The main research work is:(1)In order to improve the optimization property of the algorithm,this paper proposes a cuckoo search algorithm with dynamic step-size and discovery probability(DCS).Firstly,this paper analyzes the walking pattern of lévy flight to change the control factor of step-size of the algorithm,so that the algorithm has certain self-adaptability.After adding the step-size adjustment function,the moving distance of the individual in the population will have a nonlinear decreasing trend with the increase of iteration number of the algorithm.In this way,it accelerates the convergence performance of the algorithm.Then,it changes the discovery probability of the fixed value,and introduces the inertia weights strategy with F distribution characteristics into the discovery probability to make the value of the discovery probability randomly within a certain range.This maintains the activity of population in the algorithm,and balances the explore capability of global and local of the algorithm.Finally,this paper uses classical test functions and other algorithms to design simulation experiments for proving the characteristics of DCS algorithm to find the minimum solution.The comparative experiments demonstrate that the superiority of the DCS algorithm in these aspects of convergence speed,optimization accuracy etc.(2)In order to enhance the local search performance of the algorithm,this paper proposes the two subgroups cuckoo search algorithm based on the mean evaluation method(GCS).In the preference random walk of the basic algorithm,the algorithm uses a double random solution to generate the position of the new bird's nest,but this method is not clear and easily leads to an inefficient search of the algorithm.In order to change this situation,this paper adopts the fitness mean evaluation function to divide the current population into the better subgroup and the worse subgroup.Then,for the better subgroup,the algorithm adopts a strategy of directional mutation in the better population to generate new solutions,and makes the current individual to search in their self-direction.This avoids the phenomenon of inefficient solution due to blind searching.For the worse subgroup,the algorithm uses differential mutation mechanism of the disturbance items with the t-distribution characteristics to generate new solutions.This makes the worse individual to explore deeply in the current optimal area and enhance the excavate capability of global and local of the algorithm.The test results of the experiment show that it has better solution performance.(3)This paper uses the GCS algorithm to solve the problem of cargo distribution route planning.According to the design of the problem model,the quick sort method is used to map the individual coding methods in the population into feasible solutions to the problem.Then the local search method is used to determine the city access order,so as to solve the optimal path solution of the problem.The results of solution to the test case of the standard TSP database show that the GCS algorithm has a good adaptability and better solution effect in the problems of path planning.