CAJ | 학술논문

机构运动链同构识别是NP难问题,本文将运动链等价转换为拓扑图,运用图的同构识别原理判断运动链同构.本文高效结合遗传算法和局部搜索算法,并提出伪杂交算子,预杂交的两个体相互根据对方的元素排列信息重新对自身的元素进行排列,而不是实际的交叉,这一算子不但避免了个体中重复元素的出现,而且能够拓展搜索空间,加快收敛;用本算法与一种神经网络算法进行比较,结果证明了本算法的高效性和优越性.
궤구운동련동구식별시NP난문제,본문장운동련등개전환위탁복도,운용도적동구식별원리판단운동련동구.본문고효결합유전산법화국부수색산법,병제출위잡교산자,예잡교적량개체상호근거대방적원소배렬신식중신대자신적원소진행배렬,이불시실제적교차,저일산자불단피면료개체중중복원소적출현,이차능구탁전수색공간,가쾌수렴;용본산법여일충신경망락산법진행비교,결과증명료본산법적고효성화우월성.
A mechanism kinematic chain is uniquely represented by a topological graph. The principles of mechanism kinematic chain isomor-phism identification are proposed according to the graph theory. A new algorithm combining the genetic algorithm and local search algorithm is presented to optimize the test of isomorphism. The pseudo-crossover operator is introduced. In this operator,instead of real crossover,each unit rearranges the order of elements according to the order of elements in another unit. Necessary statistic data of experiment is present to prove the correctness and effectiveness of this hybrid algorithm.