A labor division artificial bee colony algorithm based on behavioral development

Artificial bee colony (ABC) algorithm mimics the foraging behaviour of bee colonies to solve optimization problems in which different types of bees adopt the same search equation. In this sense, bees that play different roles do not divide their labor. Moreover, the single search equation in ABC is strongly explorative but weakly exploitative, which limits its performance. To overcome that issue, this study proposes an improved algorithm (hereafter BDLDABC) that introduces the labor division of bee colonies based on behavioral development into ABC. In BDLDABC, employed bees, onlooker bees, and scout bees are regarded as three phases of the behavioral development of bees, and the labor division for the search process is obtained by individual specialization and role plasticity. According to individual specialization, three search equations guided by the global best solution, local best solution, and a random solution are designed for the three types of bees. Based on role plasticity, four patterns of behavioral development (i.e., normal development, accelerated development, delayed development, and reversed development) are designed for bees. Following the mechanics of dividing labor, bees adaptively adjust their search equations in response to changes in the search environment. Two groups of widely used benchmark functions (including fifty-two test functions) and the real-world circle packing problem are employed to verify the performance of BDLDABC, and the experimental results show that, in most cases, BDLDABC is superior, or at least comparable, to its competitors (including eight ABC variants, three DE variants, and three PSO variants).(c) 2022 Elsevier Inc. All rights reserved.

Automated summary

The ABC algorithm mimics bee foraging behavior to solve optimization problems, but its performance is limited. This study proposes an improved algorithm, BDLDABC, which introduces the concept of behavioral development and labor division based on individual specialization and role plasticity, to enhance the algorithm's performance.