An Agglomerative Greedy Brain Storm Optimization Algorithm for Solving the TSP

Changyou Wu,Xisong Fu

doi:10.1109/access.2020.3035899

Changyou Wu, Xisong Fu

Open Access

https://doi.org/10.1109/access.2020.3035899

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2020
Citations: 81	License type: CC BY 4.0

Affiliation: Shandong Institute of Business and Technology

Abstract

The brain storm optimization algorithm(BSO) is a population based metaheuristic algorithm inspried by the human conferring process that was proposed in 2010. Since its first implementation, BSO has been widely used in various fields. In this article, we propose an agglomerative greedy brain storm optimization algorithm (AG-BSO) to solve classical traveling salesman problem(TSP). Due to the low accuracy and slow convergence speed of current heuristic algorithms when solving TSP, this article consider four improvement strategies for basic BSO. First, a greedy algorithm is introduced to ensure the diversity of the population. Second, hierarchical clustering is used in place of the k-means clustering algorithm in standard BSO to eliminate the noise sensitivity of the original BSO algorithm when solving TSP. Exchange rules for the individuals in the population individuals were introduced to improve the efficiency of the algorithm. Finally, a heuristic crossover operator is used to update the individuals. In addition, the AG-BSO algorithm is compared with the genetic algorithm (GA), particle swarm optimization (PSO), the simulated annealing(SA) and the ant colony optimization (ACO) on standard TSP data sets for performance testing. We also compare it with a recently improved version of the BSO algorithm. The simulations show the encouraging results that AG-BSO greatly improved the solution accuracy, optimization speed and robustness.

Highlights

In the past two decades, swarm intelligence(SI) algorithms have become highly influential
The AG-BSO algorithm proposed in this paper converges to the optimal value in the 44th, 184th and 536th generations, respectively, while the optimal values obtained by the SA, genetic algorithm (GA), and particle swarm optimization (PSO) algorithms continue to constantly change with an increasing number of iterations
To address the above problems, this paper proposes an improved BSO algorithm named AG-BSO

Summary

INTRODUCTION

In the past two decades, swarm intelligence(SI) algorithms have become highly influential. In 2018, Dash et al [31] introduced k-means++ technology to improve the BSO algorithm, which solves the problem of the slow convergence of the algorithm by using a random probability decision in the river formation dynamics scheme to select the best clustering centroid for population generation. In 2018, Duan et al [32] introduced a new clustering method based on metric distance into the basic BSO, proposed metric distance brainstorm optimization (MDBSO), and applied the improved algorithm. A good BSO algorithm for soling TSP should have the following characteristics: (1) The individual update strategies should make full use of information on the fitness of the current population and the number of algorithm iterations.

25: Compare new and old individuals

IMPROVED BSO FOR THE TSP

SOLUTION CLUSTERING

Findings

CONCLUSION