Research On The Artificiai Bee Colony Algorithm Theories And Applications

The artificial bee colony algorithm is a relatively new swarm intelligence algorithm which has the advantages of less control parameters, fast searching speed, high searching precision and strong robustness. As a result, scholars have paid close attention to this algorithm ever since it was proposed and it has been used to solve engineering optimizing problems. In the recent years, the ABC algorithm has been widely applied to the communications, control, automation and biological engineering field, and has obtained good optimization results.However, the ABC algorithm has the following shortcomings:its population update mechanism indicates it is good at exploration but poor at exploitation; moreover, it does not use the social information effectively and lacks the knowledge of the problem structure, which leads to the insufficiency in both convergent speed and searching precision.This dissertation mainly explores the applications and the improvement mechanisms of the ABC algorithm which aim to improve the optimization performance including the convergence rate and the searching precision to further enhance the applicability of the algorithm. The main work of this dissertation including the following aspects:Firstly, we apply the ABC algorithm to the Direction of arrival (DOA) estimation problem which is one of the most important issues in array signal processing field. Experimental results show that the ABC algorithm can successfully solve this problem and get good estimation.To further explore the applicability of the ABC algorithm, we apply this algorithm to the constrained optimization problems and the multi-objective optimization problems respectively. When solving the constrained optimization problems, a new function called constraints checking function is defined to check whether the solution is in the feasible space and is used to improve the greedy criterion and the following probability to adjust the ABC algorithm to solve these problems more effectively. When comes to the multi-objective optimization problems, the fitness function and the greedy criterion is modified by the Pareto dominance principle, meanwhile, the external file mechanism is adopted to preserve and update the Pareto optimal solution. What’s more, the global-guided neighborhood searching strategy is used during the entire searching process to certify the uniformly distribution of the Pareto optimal solutions. Optimization results of the benchmark functions show the effectiveness of these two improved ABC algorithms.To revise the insufficient of the searching mechanism mentioned above, the ABC algorithm is combined with the generalized opposition-based learning (OBL) strategy and the Archimedean copula estimation distribution algorithm (ACEDA) respectively and two new hybrid algorithms are proposed in this dissertation. In the OBL based ABC algorithm, the reverse solution mechanism is adopted when updating the population to increase the diversity of the swarm. Meanwhile, we use the multidimensional neighborhood searching mechanism to improve the searching efficiency. As for the Archimedean copula EDA based ABC algorithm (ACABC), we analyses the problem structure by constructing the probabilistic model of the dominant population and generate new swarm according to the estimation distribution information to achieve directed optimization. Simulation results indicate that both hybrid algorithms enhanced the searching precision and optimization efficiency obviously.