基于协同进化的多目标优化算法研究及应用

发布时间：2018-06-13 12:53

本文选题：协同进化 + 多目标优化　；参考：《南京邮电大学》2017年硕士论文

【摘要】：工程实践与科学研究中会经常遇到一些多目标优化问题,这些优化问题如果采用传统的解决方法处理会出现一定的缺陷和弊端。虽然使用进化算法解决多目标优化问题已被证明是一个有效的方法,但是目前进化多目标优化算法还不完善,这些进化算法还存在解集分布不均匀、收敛早熟和精度差等缺点。协同进化算法是近年来提出的一种解决多目标优化问题的新方法,与传统进化多目标优化算法相比,它能在一定的程度上提高全局收敛性和避免早熟。然而目前的协同进化多目标优化算法在解决多目标问题,然而目前的协同进化多目标优化算法在解决多目标问题,得出非支配解集的分布性和多样性方面不太理想,并且算法的全局收敛性还有待进一步提高。本文针对目前协同进化算法的不足提出了相应的改进措施和策略,整合改进的协同进化算法和进化多目标优化机制,研究出更有效的协同进化多目标优化算法,并将该算法有效地应用于解决机器人多目标路径规划问题,主要研究工作如下:(1)针对协同进化算法中选取代表个体的导向性不强的问题,提出了一种分组排序评估策略的合作协同进化算法。通过不断将每一代新种群进行有序排列的分组评估,选择当代最优个体组成代表组合,使选取的代表组合更具有导向性。将该算法与其它进化算法采用典型的测试函数进行对比测试实验,结果表明改进的协同进化算法具有更快的收敛速度和更强的全局收敛能力。将上述提出的基于分组排序评估策略的合作协同进化算法应用于复杂PID控制系统参数优化,实验结果表明,该算法能高效地搜索到给定性能指标要求的PID参数最优组合,具有较好的应用前景。(2)针对多目标优化算法中非支配解空间分布不均匀以及算法收敛精度不高的问题,将多种群协作的思想、快速非支配排序的方法以及精英外部档案的策略相结合,提出了一种多种群合作协同多目标优化算法。采用标准的多目标优化问题测试函数集对所提出的算法与NSGA-II算法进行对比测试实验,结果表明,所提出的多种群合作协同多目标优化算法,能获得更均匀和更精确的非支配解集,达到更优的Pareto前沿面。(3)研究了应用所提出的多种群合作协同多目标优化算法解决机器人多目标路径规划问题的方法及其实现。将机器人多目标路径规划任务进行建模,提出了其包含多项性能指标要求的多目标优化模型;给出了多种群合作协同多目标优化算法求解的实现方法。仿真实验结果表明,所提出的方法能有效地获得多目标要求下的优化路径。
[Abstract]:Some multi-objective optimization problems are often encountered in engineering practice and scientific research. If traditional methods are adopted to solve these optimization problems, there will be some defects and drawbacks. Although using evolutionary algorithm to solve multi-objective optimization problem has been proved to be an effective method, but at present evolutionary multi-objective optimization algorithm is not perfect, these evolutionary algorithms still have some shortcomings, such as uneven distribution of solution set, premature convergence and poor precision. Coevolutionary algorithm is a new method to solve the multi-objective optimization problem proposed in recent years. Compared with the traditional evolutionary multi-objective optimization algorithm, it can improve the global convergence and avoid prematurity to a certain extent. However, the current coevolutionary multi-objective optimization algorithm is not ideal in solving the multi-objective problem and obtaining the distribution and diversity of the non-dominated solution set. And the global convergence of the algorithm needs to be further improved. In this paper, the corresponding improvement measures and strategies are put forward to overcome the shortcomings of the current co-evolution algorithm, which integrates the improved co-evolution algorithm and the evolutionary multi-objective optimization mechanism, and develops a more effective co-evolution multi-objective optimization algorithm. The algorithm is effectively applied to solve the multi-objective path planning problem of robot. The main research work is as follows: (1) aiming at the problem that the selection of representative individual is not strong in the co-evolutionary algorithm, the main research work is as follows: A cooperative coevolutionary algorithm for grouping ranking evaluation strategy is proposed. Through the grouping evaluation of each generation of new population in order, the representative combination of the best individual is selected, which makes the representative combination more oriented. The experimental results show that the improved co-evolutionary algorithm has faster convergence speed and stronger global convergence ability than other evolutionary algorithms. The cooperative coevolutionary algorithm based on grouping ranking evaluation strategy is applied to the parameter optimization of complex pid control system. The experimental results show that the algorithm can efficiently search the optimal combination of pid parameters required by given performance index. It has a good application prospect. (2) aiming at the problems of non-dominated solution spatial distribution and low convergence precision of multi-objective optimization algorithm, the idea of multi-group cooperation is put forward. A multi-group cooperative and multi-objective optimization algorithm is proposed based on the combination of the fast non-dominated sorting method and the strategy of elite external files. The standard multi-objective optimization problem test function set is used to compare the proposed algorithm with NSGA-II algorithm. The results show that the multi-group cooperative multi-objective optimization algorithm is proposed. A more uniform and accurate set of non-dominated solutions is obtained to achieve a more optimal Pareto frontier. The method and implementation of multi-group cooperative multi-objective optimization algorithm for robot multi-objective path planning are studied in this paper. The multi-objective path planning task of robot is modeled, and the multi-objective optimization model with multiple performance requirements is proposed, and the implementation method of multi-group cooperative multi-objective optimization algorithm is presented. The simulation results show that the proposed method can effectively obtain the optimal path under multi-objective requirements.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】