中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
遗传算法及其在机器人技术中的应用

文献类型:学位论文

作者刘雁飞
学位类别工学硕士
答辩日期1999-06-01
授予单位中国科学院自动化研究所
授予地点中国科学院自动化研究所
导师裘聿皇
关键词遗传算法 两层编码 机械手控制 路径规划 genetic algorithms two-layer encoding manipulator control robot path planning
学位专业控制理论与控制工程
中文摘要本文对遗传算法在理论分析方面进行了归纳总结,并在此基础上从分子遗 传学中获得灵感,提出了一种新的编码方法一两层编码。共进行了三方面的工 作:两层编码的遗传算法、两层编码的遗传算法与分割空削法相结合在移动机 器人路径规划问题中的应用和先"自交"后"杂交"的遗传算法在机械手控制 中的应用。 我们从分子遗传学中遗传信息的表达过程,即 碱基A,C,G,U 氩尽酸 蛋白质 中发现这种两层表达比起一层表达来更能清晰、明了地说明蛋白质的组成,氨 基酸这一中间表达使得在研究蛋白质是如何体现生物性状这一问题上如鱼得 水。编码就是一个信息表达的过程,于是我们汲取了遗传信息表达过程的精髓, 提出了新的编码方法一两层编码。在第二章中我们对它进行了详细的阐述,并 在第三章中把它应用到机器人的路径规划问题中。 我们知道用分割单元法解决移动机器人路径规划问题时,最优路径是由一 连串的单元组成的。本文提出了一种新的分割单元的方法,并在此基础上运用 两层编码的遗传算法进行搜索。用遗传算法来解决这个问题时,这一连串的单 元组成就是编码,我们发现很多不同的编码其实是对应同一个代价,也就是适 应度函数值。为了提高搜索效率,我们运用了两层编码,从→,↑和 到由这 三个方向组成的1 0个不同的到达位置("表现型")为第一层,从这10种 "表现型"到由这些"表现型"组成的最后路径为第二层。仿真实验结果说明 了这种两层编码方法的遗传算法和分割单元法相结合的算法能够降低搜索的复 杂度。 机械手控制问题因为涉及到求雅可比矩阵的逆,一直是一个难解的非线性 问题,尤其是对非冗余度机器人。近几年来,人们提出了几种基于神经网络和 模拟退火算法的方法,这些方法中,有的过于繁琐,有的控制精度不高。本文 引入遗传算法,试图用一种新的思路来解决这个问题。我们对遗传算法进行了 改进,把分区间段先"自交"后"杂交"的遗传算法应用到机械手控制中,在 实例平面两白山度机械手上,取得了较高的控制精度,并与其它方法的结果逃 行了比较。
英文摘要In this dissertation, we make a summary of genetic algorithms ( GAs ) in theory analysis. We propose a new kind of encoding approach for the GAs, called two-layer encoding, which comes from the genetic mechanism in molecular genetics. Three main problems have been studied, which are GAs with two-layer encoding, mobile robot path planning based on GAs with two-layer encoding and GAs used in manipulator control. From the process of genetic information expression: Bases A,C,G and U amino acids protein we find that the two-layer expression can express the composition of protein more clearly than one-layer expression. Encoding is a process of information expression. So we grasp the key of genetic information expression and propose a new kind of encoding: two-layer encoding. In chapter two we describe this encoding approach in details. In chapter three we apply it into mobile robot path planning problem. When we apply cell decomposition method in mobile robot path planning, the optimal path is composed of a series of cells. We present a new kind of cell decomposition method, which has been applied into this path planning problem combined with GAs with two-layer encoding. In this problem, a series of cells are codes in GAs. We find that many different codes lead to one cost, i.e. the value of fitness function. In order to improve the efficiency of exploration, we use two-layer encoding. The first layer encoding is from →,↑and to ten projections ("phenotype") reaching to ten points through different composition of →, ↑ and . The second layer is from ten kinds of "phenotype" to the final path composed of such "phenotype". Experiment results have proved that this algorithm decreased the exploration complexity. Manipulator control is a kind of complex nonlinear problem, because it relates to the problem of solving the inverse of Jacobian matrix, especially to redundant robots. In these years, some methods based on neural network and simulated annealing have been presented. We propose a new kind of GAs, named first "autocopulation" then "hybridization" GAs, and apply it to a planar manipulator with two degrees of freedom. Simulation results have proved that this method make high control accuracy compared with other methods.
语种中文
其他标识符501
源URL[http://ir.ia.ac.cn/handle/173211/7247]  
专题毕业生_硕士学位论文
推荐引用方式
GB/T 7714
刘雁飞. 遗传算法及其在机器人技术中的应用[D]. 中国科学院自动化研究所. 中国科学院自动化研究所. 1999.

入库方式: OAI收割

来源:自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。