An Improved Differential Evolutionary Algorithm And Its Application In Wireless Sensor Network Localization Problem

In real life, many engineering problems can be converted to global optimization problems.The goal of global optimization is to accurately and rapidly search the optimum solution in feasible regions. Since the global optimization problem has been proposed, a large number of scholars have conducted deeply research. Many methods have been proposed, which can be divided into the deterministic and random methods. The deterministic optimization methods commonly include gradient descent method, Newton method, conjugate gradient method, the simplex method and so on. The significant advantages of such methods are fast convergence. But there are some assumptions for the objective functions, such as continuity,differentiability and so on. Because of fixed search direction, these methods hardly jump out the local optimization. With more and more high-dimensional complex optimization problems, people pay more attention to random methods. Due to without any assumptions for objective functions, the random methods can be used to solve any optimization problems.If the reasonable search process was designed, they can accurately find the global minimum point.Differential evolutionary algorithm is a new branch of evolutionary algorithm. Since its structure is simple, robust and easy to combine with other methods, it has been widely used in various research fields. However, the existing differential evolutionary algorithms are weakly to break out the local optimization and not effectively use the objective functions’ properties. These deficiencies seriously affect the performance of differential evolutionary algorithms and hinder its further applied to our life.In this thesis, differential evolutionary algorithm and its application are investigated. First, differential evolutionary algorithm with strategy adaptation for global numerical optimization is introduced. But its ability of global search performance is weak and does not use the properties of objective functions. For the above shortcomings, adaptive differential evolutionary algorithm based on hyperspherical coordinates and gradient search strategies is proposed. In order to verify the algorithm effectiveness for the high-dimensional optimization problems,it is tested on 14 classic test functions and compared the results with other algorithms. To be able to further applied the differential evolutionary algorithm to actual life, the differential evolutionary algorithm is used to solve the wireless sensor network localization problem.After analyzing the general model, this thesis finds that its assumption for sensor connecting radius is deficient and the stability is not considered. Therefore, the improved wireless sensor networks positioning model is proposed to solve limitations. Subsequently, according to the characteristics of localization problems, adaptive differential evolutionary algorithm based on hyperspherical coordinates and gradient search strategies is optimized. The optimized algorithm has a stronger global search ability at early stage and its local search ability is prominent in the later. Finally, using computers simulating the wireless sensor network,two groups of experiments are used to test the algorithm, which prove the effectiveness of the algorithm for wireless sensor network optimization problem.