不确定区域下基于自适应并行遗传的最优鲁棒控制器

Optimal robust controller based on adaptive parallel genetic algorithms in uncertain domain

中文关键词  不确定区域  最优控制  并行遗传  鲁棒稳定  概率统计

英文关键词  uncertain domain  optimal control  parallel genetic algorithm  robust stability  probability and statistics

中文摘要

      目前使用遗传算法设计鲁棒控制器时, 都要人为地给定变量搜索空间. 当变量区域不确定时, 采用自适应并行遗传算法设计出最优鲁棒控制器, 该方法根据当前搜索到的各种群最优个体的分布情况, 运用概率统计理论求出变量区域的最小方差无偏估计, 不断缩小不确定的变量搜索区域, 从而逐步达到最优, 并且考虑了种群个体适应度对算法中交叉概率和变异概率的影响. 该方法能设计出简单正则、阶数低的最优鲁棒控制器, 而且仿真结果表明, 这些控制器有效地避免了局部最优, 提高了算法的寻优精度和收敛速度.

英文摘要

      In most of the existing robust controller designs based on the genetic algorithm, the search domains of variable are factitiously assigned. For undefined search domains, we propose a novel method for designing the optimal robust controller based on the adaptive parallel genetic algorithm. According to the distribution of the best individuals of each population in the current search, the algorithm narrows the uncertain search domain of variables by using the probability and statistics theory, so as to gradually achieve the optimum. Moreover, the impact of the individual fitness on the crossover probability and mutation probability is also investigated. By using this method, we design a simple, proper and low-order optimal robust controller, which effectively avoids the local optimum and improves the convergence and accuracy in simulation experiments.