Research Of Adjustable And Dynamic Artificial Bee Colony Algorithm

Swarm intelligence refers to the collective intelligent behavior through the interactions of the individuals among self-organized swarms. The problems that the natural intelligent swarms can solve are extremely similar to many engineering areas of real world. Swarm intelligence has become the important research area of computer science, economics, biology, engineering and many other disciplines.Artificial Bee Colony algorithm (ABC) is a relatively new branch optimization method of swarm intelligence. In recent years, the ABC algorithm has gradually become more popular due to its simplicity, less control parameters and ease of implementation. However, just like any other stochastic searching algorithm, ABC algorithm tends to get trapped at the local minima when solving some complex optimization problems. In order to improve the performance and effectiveness of the ABC algorithm, this article proposed some improvements from different perspectives.Our work is primarily around Artificial Bee Colony algorithm. Firstly, this paper summarizes the current situation on ABC algorithm, and the biology background. The mathematical model and the algorithm process of ABC algorithm are discussed in detail.Secondly, this paper proposes an improved algorithm called Improved Artificial Bee Colony algorithm with Self-adaptive step length (SA-IABC).The basic idea is adjusting step size adaptively to control the range of neighborhood during the process of searching. Exploiting in the small search space can improve the optimization precision, while exploring in the large search space can improve the speed of convergence. Algorithm carries out exploiting and exploring at the same time to balance the conflict of convergence and precision, and avoid stagnation phenomenon. In addition, a different probability selection formula is introduced to maintain the diversity of population. The results of simulation experiment show that SA-IABC has superior convergence and the ability to avoid premature convergence. And it is proven to be especially suitable for optimizing multi-modal problems.Finally, a novel Dynamic Artificial Bee Colony Based on Chaos (DABCBC) which combined with chaotic search method is proposed. The new algorithm makes good use of the stochastic, ergodicity and regularity property rather than a random number generator. In the initialization, the algorithm generates chaotic variables as the initial solutions, while in the later phase of scouts, a good solution is generated using chaotic search to get rid of local minima. Besides, tournament selection strategy in the process of onlookers’selecting food sources is put to use. The experimental results indicate that DABCBC not only guarantees population diversity, avoids getting trapped at local minima, but also increases the speed of evolution. And It can solve high-dimensional complex problems effectively.