基于突变策略的自适应骨干粒子群算法

A Self-Adaptive Bare-Bones Particle Swarm Optimization Algorithm Based on Mutation Strategy

摘要: 骨干粒子群算法是由标准粒子群算法演变而来的，其在粒子位置更新方面采用了高斯采样策略。针对骨干粒子群算法在解决高维优化问题时存在的易陷入局部最优的问题，文中引入了具有下降趋势的时变因子，提出了一种基于突变策略的带有自适应扰动值的骨干粒子群算法。该算法在高斯分布的均值项中引入两个服从均匀分布的随机数，在高斯分布的标准差中引入了一个自适应扰动值，且给出了突变策略进一步保证粒子收敛到全局最优解。改进后的算法与其他5种粒子群算法在9个经典测试函数上进行仿真实验，结果表明改进的算法在收敛速度和收敛精度方面的综合表现都优于其它算法。

Abstract: The bare-bones particle swarm optimization algorithm is evolved from the standard particle swarm optimization algorithm, which adopts the Gaussian sampling strategy when particles’ position update. To improve the problem of premature convergence when solving the high-dimensional optimization problems, a time-varying factor with downward trend is introduced and a self-adaptive mutation bare-bones particle swarm optimization (AMBPSO) is proposed where the particle would mutate according to adaptive probability when it becomes stagnant after updating with the perturbation strategy in order to jump out of the local optimum. The algorithm introduces two random numbers that obey uniform distribution in the mean term of the Gaussian distribution and an adaptive perturbation value in the standard deviation of the Gaussian distribution, and gives a mutation strategy to further ensure that the particles converge to the global optimum. The proposed algorithm and other five particle swarm optimization algorithms are simulated on nine classical benchmark functions, whose experimental results show that the proposed AMBPSO algo-rithm is more competitive than some existing popular variants of PSO algorithms in terms of con-vergence speed and convergence accuracy.