基于树木枝梢生长分枝过程的仿生计算算法

Bionic computational algorithm based on growth branching process of trees shoot

借鉴树木枝梢向光性生长的形态特征,提出一种仿生智能优化算法——生长树算法。将树木具有生长发育的器官(主干和枝叶)抽象地定义为树木的枝梢个体,将其映射为生长树算法的编码,新一层次枝梢个体的编码是上一层次枝梢个体编码的继承和扩展,通过判断每个可能的新一层次枝梢个体的生长素浓度的大小,选择其中生长素浓度较大的个体生成下一层次的枝梢种群,形成整个树木,即生长树,当生长树达到一定的层次时,加入枝梢个体编码的变形机制,增强枝梢个体在生长树生长空间的搜索能力,使算法趋近于最优解。算法由3个阶段组成:快速发育阶段、成长阶段、衰老阶段。通过利用经典的优化函数进行寻优实验,与遗传算法和蚁群算法的寻优结果进行对比,实验结果表明,该算法迭代次数少、收敛速度快、求解精度高,相比遗传算法和蚁群算法有更好的寻优能力。

A bionic intelligent optimization algorithm was proposed refering on the phototropic growth of morphology of trees shoot. The algorithm called growth trees algorithm. The trunk and foliage of the trees were abstractly defined as the individual shoots of the trees and then mapped into the tree of the growth tree. The coding of the new level shoots was the coding of the individual inheritance and expansion, by determining the size of the concentration of auxin of each individual new level of shoots, the larger concentration of auxin in the next generation of individuals was selected to produce the next level of shoots to form the entire tree, namely growth tree, when the growth tree reached a certain level, the deformation mechanism of individual shoots was added to enhance the searching ability of the individual growth space in the growth tree, which made the algorithm approach the optimal solution. The algorithm consisted of three phases including rapid development stage, growth stage and senescence. Growth trees algorithm and genetic algorithm were compared through classic function optimization problems. The results show that the growth trees algorithm has higher accuracy, faster convergence and fewer iterations. It has better optimization capability than genetic algorithm and ant colony algorithm.