A Hybrid Genetic Algorithm for Bin Packing Problem Based on Item Sequencing
一种求解装箱问题的混合算法-最新文档
一种求解装箱问题的混合算法: This paper presents a new hybrid algorithm based on genetic algorithm and tabu search for bin packing problem. It combines the advantage of global search ability of genetic algorithm with the adaptability of tabu search and has better convergence performance than simple genetic algorithm. At last, an practical example is applied to prove the efficiency of this algorithm.0引言装箱问题(Bin Packing Problem, BPP)是一类重要的组合优化问题,在现实生活中有着广泛的应用背景,特别在现代物流中,许多问题都抽象化为装箱问题或其变形,如货物如何装载,才能提高运载器具的利用率,从而降低运输成本;物流任务应如何调度,才能提高运行效率,等等。
但在理论上,装箱问题是一个NP难题[1],很难精确求解。
因此对其求解进行研究具有重要的理论价值和实际意义。
到目前为止,针对该问题人们提出了许多算法,但都有其局限性:枚举法和分支定界等精确算法在箱子数目稍大时,会出现“组合爆炸”;一些近似算法如下次适应NF、首次适应FF、降序首次适应FFD、最佳适应BF等,在解决复杂的装箱问题时结果与物品的体积数据有较大关系,在极端情况下很不理想;遗传算法能在合理的时间内求得最优解或满意解,但易陷入局部最优。
本文针对以上算法的不足,提出一种混合算法,该算法结合遗传算法良好的全局搜索能力和禁忌搜索具有记忆能力的全局逐步优化特性,增强全局和局部意义下的搜索能力和效率。
《供应商管理研究国内外文献综述1800字》
供应商管理研究国内外文献综述Dickson G.W(1966)作为较早提出了供应商绩效管理这一概念的学者。
早在1966年,该学者通过问卷及访谈的方法对采购员进行了调查,结果表明不同的采购商、不同的行业等可以总结出超过55项指标来对供应商进行评价,进一步分析后他认为其中23项指标可以作为基础标准。
后续的研究中不断有学者对供应商评价进行了深入的研究,让整个理论越来越丰富。
R Kant(2017)在其研究中针对前人的研究进行了补充,在其大量的研究资料中都不断丰富Dickson的相关研究,对其提出的23项评价标准进行了侧面探讨,经过他的研究后发现供应商评价和选择的核心准则可以分别概括为货品质量、交货速度和价格三个方面。
后续其他学者的研究可以总结为如下几个层面的供应商评价指标体系:(1)广泛意义上的供应商评价指标研究MN Mohd Nawi(2016)在其研究中在前人研究基础上,发现价格、质量、服务、技术等10个指标相对较为重要,尤其是针对一些制造产业来说,价格对小企业的影响权重更大。
Ferhan Çebi(2016)在其研究中从政府供应商的角度入手,他发现政府等公办机构选择供应商的过程中更倾向于将质量和供应商实力作为重要评价指标。
张敏(2017)在其研究中引入了精益供应链这一概念,后续通过调查问卷、访谈等研究,确定了一套质量、交货速度、价格、供应商发展潜力、合作效率、供应商财务健康状况、供应商外部环境评价等指标为一体的评价指标体系,并且定义为精益供应链供应商。
黄虹(2017)在其研究中引入了战略供应商这一概念,战略合作伙伴不同于一般意义的供应商合作伙伴,更是要将供应商与生产商之间形成严格合规的体系,所以他认为应当从合作兼容性、成本、交货速度、产品质量、发展能力等方面对供应商进行评价,后续的研究中他认为生产商组织能力和战略合作能力也是重要的评价因素。
(2)特殊产业的供应商评价指标研究不同产业对于供应链的评价指标体系有不同的细节需求,一些特殊行业的供应商评价指标研究能够给照明产业一定的启示。
遗传算法优化svm参数
遗传算法优化svm参数遗传算法是一种基于自然适应性进化理论的优化算法,它通过模拟自然界中的进化过程,通过遗传算子(交叉和变异操作)对个体进行进化和选择,以找到最优解决方案。
支持向量机(Support Vector Machine,SVM)是一种非常有效的分类算法,通过在数据集中找到最有代表性的样本点,构建超平面分离不同类别的样本。
优化SVM的参数可以提高分类的准确率和稳定性。
下面是使用遗传算法优化SVM参数的一般步骤:1. 确定优化目标:首先明确需要优化的SVM参数,如惩罚系数C、核函数类型和参数、松弛变量等,这些参数会影响模型的性能。
2. 设计基因编码:将待优化的参数映射为基因的编码形式,可以使用二进制、整数或浮点数编码。
例如,某个参数的取值范围为[0, 1],可以使用浮点数编码。
3. 初始化种群:随机生成初始的种群,每个个体都表示一个SVM参数的取值组合。
4. 适应度评估:使用训练集对每个个体进行评估,计算其在测试集上的准确率或其他指标作为个体的适应度。
5. 选择操作:根据适应度排序或轮盘赌等策略,选择优秀个体进行遗传操作。
6. 交叉操作:从选中的个体中进行交叉操作,生成新的个体。
可以使用单点交叉、多点交叉或均匀交叉等策略。
7. 变异操作:对生成的新个体进行变异操作,引入随机扰动,增加种群的多样性。
变异操作可以改变某个基因的值或重新随机生成某个基因。
8. 更新种群:将交叉和变异生成的个体合并到种群中。
9. 重复步骤4-8,直到满足终止条件(如达到最大迭代次数或种群适应度不再改变)。
10. 选择最优个体:从最终的种群中选择适应度最好的个体作为最优解,即SVM的最优参数。
通过以上步骤,遗传算法可以搜索参数空间,并找到最有解决方案。
通过尝试不同的参数组合,可以优化SVM模型的性能。
请注意,以上只是一般的遗传算法优化SVM参数的步骤,实际应用中可能会根据具体问题进行适当的调整。
在实际操作中,还可以通过引入其他优化技巧(如局部搜索)来进一步提高搜索效率。
毕业设计论文基于遗传算法的BP神经网络的优化问题研究.doc
编号:审定成绩:重庆邮电大学毕业设计(论文)设计(论文)题目:基于遗传算法的BP神经网络的优化问题研究学院名称:学生姓名:专业:班级:学号:指导教师:答辩组负责人:填表时间:2010年06月重庆邮电大学教务处制摘要本文的主要研究工作如下:1、介绍了遗传算法的起源、发展和应用,阐述了遗传算法的基本操作,基本原理和遗传算法的特点。
2、介绍了人工神经网络的发展,基本原理,BP神经网络的结构以及BP算法。
3、利用遗传算法全局搜索能力强的特点与人工神经网络模型学习能力强的特点,把遗传算法用于神经网络初始权重的优化,设计出混合GA-BP算法,可以在一定程度上克服神经网络模型训练中普遍存在的局部极小点问题。
4、对某型导弹测试设备故障诊断建立神经网络,用GA直接训练BP神经网络权值,然后与纯BP算法相比较。
再用改进的GA-BP算法进行神经网络训练和检验,运用Matlab软件进行仿真,结果表明,用改进的GA-BP算法优化神经网络无论从收敛速度、误差及精度都明显高于未进行优化的BP神经网络,将两者结合从而得到比现有学习算法更好的学习效果。
【关键词】神经网络BP算法遗传算法ABSTRACTThe main research work is as follows:1. Describing the origin of the genetic algorithm, development and application, explain the basic operations of genetic algorithm, the basic principles and characteristics of genetic algorithms.2. Describing the development of artificial neural network, the basic principle, BP neural network structure and BP.3. Using the genetic algorithm global search capability of the characteristics and learning ability of artificial neural network model with strong features, the genetic algorithm for neural network initial weights of the optimization, design hybrid GA-BP algorithm, to a certain extent, overcome nerves ubiquitous network model training local minimum problem.4. A missile test on the fault diagnosis of neural network, trained with the GA directly to BP neural network weights, and then compared with the pure BP algorithm. Then the improved GA-BP algorithm neural network training and testing, use of Matlab software simulation results show that the improved GA-BP algorithm to optimize neural network in terms of convergence rate, error and accuracy were significantly higher than optimized BP neural network, a combination of both to be better than existing learning algorithm learning.Key words:neural network back-propagation algorithms genetic algorithms目录第一章绪论 (1)1.1 遗传算法的起源 (1)1.2 遗传算法的发展和应用 (1)1.2.1 遗传算法的发展过程 (1)1.2.2 遗传算法的应用领域 (2)1.3 基于遗传算法的BP神经网络 (3)1.4 本章小结 (4)第二章遗传算法 (5)2.1 遗传算法基本操作 (5)2.1.1 选择(Selection) (5)2.1.2 交叉(Crossover) (6)2.1.3 变异(Mutation) (7)2.2 遗传算法基本思想 (8)2.3 遗传算法的特点 (9)2.3.1 常规的寻优算法 (9)2.3.2 遗传算法与常规寻优算法的比较 (10)2.4 本章小结 (11)第三章神经网络 (12)3.1 人工神经网络发展 (12)3.2 神经网络基本原理 (12)3.2.1 神经元模型 (12)3.2.2 神经网络结构及工作方式 (14)3.2.3 神经网络原理概要 (15)3.3 BP神经网络 (15)3.4 本章小结 (21)第四章遗传算法优化BP神经网络 (22)4.1 遗传算法优化神经网络概述 (22)4.1.1 用遗传算法优化神经网络结构 (22)4.1.2 用遗传算法优化神经网络连接权值 (22)4.2 GA-BP优化方案及算法实现 (23)4.3 GA-BP仿真实现 (24)4.3.1 用GA直接训练BP网络的权值算法 (25)4.3.2 纯BP算法 (26)4.3.3 GA训练BP网络的权值与纯BP算法的比较 (28)4.3.4 混合GA-BP算法 (28)4.4 本章小结 (31)结论 (32)致谢 (33)参考文献 (34)附录 (35)1 英文原文 (35)2 英文翻译 (42)3 源程序 (47)第一章绪论1.1 遗传算法的起源从生物学上看,生物个体是由细胞组成的,而细胞则主要由细胞膜、细胞质、和细胞核构成。
遗传算法与模拟退火算法比较
一、遗传算法与模拟退火算法比较分析模拟退火算法的基本原理可以看出,模拟退火算法是通过温度的不断下降渐进产生出最优解的过程,是一个列马尔科夫链序列,在一定温度下不断重复Metropolis过程,目标函数值满足Boltzmann概率分布。
在温度下降足够慢的条件下,Boltzmann分布收敛于全局最小状态的均匀分布,从而保证模拟退火算法以概率为1收敛到全局最优。
另外,不难看出,模拟退火算法还存在计算结构简单、通用性好以及鲁棒性强等优点。
但是,模拟退火算法存在如下缺陷:1. 尽管温度参数下降缓慢时理论上可以保证算法以概率为1地收敛到最优值,但是需要的时间过长加之误差积累与时间长度的限制,难以保证计算结果为最优;2.如果降温过程加快,很可能得不到全局最优解,因此,温度的控制是一个需要解决的问题;3.在每一种温度下什么时候系统达到平衡状态,即需要多少次Metropolis过程不易把握,从而影响模拟退火算法的最终结果。
与模拟退火算法相比较,遗传算法具有如下典型特征:这两种算法的相同点是都采用进化控制优化的过程。
主要不同点是模拟退火是采用单个个体进行进化,遗传算法是采用种群进行进化。
模拟退火一般新解优于当前解才接受新解,并且还需要通过温度参数进行选择,并通过变异操作产生新个体。
而遗传算法新解是通过选择操作进行选择个体,并通过交叉和变异产生新个体。
具体说来,遗传算法具有如下特点:(1)与自然界相似,遗传算法对求解问题的本身一无所知,对搜索空间没有任何要求(如函数可导、光滑性、连通性等),只以决策编码变量作为运算对象并对算法所产生的染色体进行评价,可用于求解无数值概念或很难有数值概念的优化问题,应用范围广泛;(2)搜索过程不直接作用到变量上,直接对参数集进行编码操作,操作对象可以是集合、序列、矩阵、树、图、链和表等;(3)搜索过程是一组解迭代到另一组解,采用同时处理群体中多个个体的方法,因此,算法具有并行特性;(4)遗传算法利用概率转移规则,可以在一个具有不确定性的空间寻优,与一般的随机性优化方法相比,它不是从一点出发按照一条固定路线寻优,而是在整个可行解空间同时搜索,可以有效避免陷入局部极值点,具有全局最优特性;(5)遗传算法有很强的容错能力.由于遗传算法初始解是一个种群,通过选择、交叉、变异等操作能够迅速排除与最优解相差较大的劣解.与模拟退火算法相比,遗传算法存在局部搜索能力差、容易陷入过早收敛等缺陷,因此,人们将模拟退火算法与遗传算法相结合得到的混合算法可以避免两种算法的缺陷,有利于丰富优化过程的搜索行为,增强全局和局部意义下的搜索能力和效率。
基于预测的遗传算法
八石)=蠢 称Sigmoid激发函数: 4一,
(1)
I十C
输出神经元的激发函数为:
厂(算)=kx
(2)
4 实验研究 为r验证上一节中所提出的NeuroPGA算法的有效性,
本节利用该算法优化8个典型的测试函数,并将其优化结果 与常规GA的优化结果相比较。 4.1 测试函数
8个测试函数如下,除了Fl是最大值问题,其它函数均 为最小值问题。其中大多数函数为多模态函数:
二维矩形条带装箱问题的底部左齐择优匹配算法
ISSN 1000-9825, CODEN RUXUEW E-mail: jos@Journal of Software, Vol.20, No.6, June 2009, pp.1528−1538 doi: 10.3724/SP.J.1001.2009.03395 Tel/Fax: +86-10-62562563© by Institute of Software, the Chinese Academy of Sciences. All rights reserved.∗二维矩形条带装箱问题的底部左齐择优匹配算法蒋兴波1,2, 吕肖庆1,3+, 刘成城11(北京大学计算机科学技术研究所,北京 100871)2(第二军医大学卫生勤务学系,上海 200433)3(北京大学电子出版新技术国家工程研究中心,北京 100871)Lowest-Level Left Align Best-Fit Algorithm for the 2D Rectangular Strip Packing ProblemJIANG Xing-Bo1,2, LÜ Xiao-Qing1,3+, LIU Cheng-Cheng11(Institute of Computer Science and Technology, Peking University, Beijing 100871, China)2(Faculty of Health Services, Second Military Medical University, Shanghai 200433, China)3(National Engineering Research Center of New Technology in Electronic Publishing, Peking University, Beijing 100871, China)+ Corresponding author: E-mail: lvxiaoqing@Jiang XB, Lü XQ, Liu CC. Lowest-Level left align best-fit algorithm for the 2D rectangular strip packingproblem. Journal of Software, 2009,20(6):1528−1538. /1000-9825/3395.htmAbstract: In this paper, a heuristic placement algorithm for the two-dimensional rectangular strip packingproblem, lowest-level left align best fit (LLABF) algorithm, is presented. The LLABF algorithm is based on thebest-fit priority principle with overall consideration of several heuristic rules, such as full-fit first rule, width-fit firstrule, height-fit first rule, joint-width-fit first rule and placeable first rule. Unlike the bottom-left (BL), theimproved-bottom-left (IBL) and the bottom-left-fill (BLF) heuristic placement algorithms, LLABF algorithmdynamically selects the best-fit rectangle for packing. The computation result shows that 2DR-SPP can be solvedmore effectively by combining the LLABF algorithm with the genetic algorithm (GA).Key words: lowest-level left align best fit (LLABF) algorithm; genetic algorithm; 2D rectangular strip packingproblem; heuristic placement algorithm摘要: 针对二维矩形条带装箱问题提出了一种启发式布局算法,即底部左齐择优匹配算法(lowest-level left alignbest fit,简称LLABF). LLABF算法遵循最佳匹配优先原则,该原则综合考虑完全匹配优先、宽度匹配优先、高度匹配优先、组合宽度匹配优先及可装入优先等启发式规则.与BL(bottom-left),IBL(improved-bottom-left)与BLF(bottom-left-fill)等启发算法不同的是,LLABF能够在矩形装入过程中自动选择与可装区域匹配的下一个待装矩形.计算结果表明,LLABF结合遗传算法(genetic algorithm,简称GA)解决二维条带装箱问题更加有效.关键词: 最低左对齐最佳匹配(LLABF)算法;遗传算法;二维矩形条带装箱问题;启发式布局算法中图法分类号: TP18文献标识码: A二维矩形条带装箱问题(2D rectangular strip packing problem,简称2DR-SPP)是一个典型的组合优化问题,∗ Received 2008-02-29; Revised 2008-04-15; Accepted 2008-06-03蒋兴波等:二维矩形条带装箱问题的底部左齐择优匹配算法1529其应用相当广泛,如工业领域中的新闻组版、布料切割、金属下料等.2DR-SPP通常是指将若干个不同规格的矩形{π1,π2,…,πn}装入宽度W固定、长度L不限的矩形容器C中,要求装完所有矩形后占用高度H packing最小,从而达到节省材料的目的.在矩形的装入过程中,要求满足:①πi,πj互不重叠,i≠j,i,j=1,2,…,n;②πi必须装入在矩形容器C内,即矩形在装入过程中不能超出容器C的宽度W;③πi的边必须与矩形容器C的边平行,πi可以90°旋转.2DR-SPP是一个NP完全问题,其计算复杂度随着问题规模的增大呈指数增长.针对该类问题,国内外相关学者作了大量的研究.Hifi[1]给出了一种基于分支定界方法(branch-and-bound procedure)的精确算法,适用于解决中小规模的2D装箱问题;Lesh等人[2]针对2D矩形完美装箱问题,提出了基于分支定界的穷举搜索算法,该算法已证明对于低于30个矩形的装箱是有效的.为了解决大规模的矩形装箱问题,一些启发式算法也相继提出来.Zhang等人[3]提出了一种启发式递归算法HR(heuristic recursive algorithm),该方法基于启发式策略和递归结构,实验结果表明,该算法能够在较短时间内获得较为理想的装箱高度,但其平均运行时间却达到了O(n3).陈端兵等人[4]根据先占角后占边的原则,提出了一种针对矩形装箱的贪心算法.Chen等人[5]给出了一种two-level搜索算法来求解2DR-SPP.Cui[6]给出了一种启发式递归分支定界算法HRBB(heuristic recursive branch-and-bound algorithm),该算法将递归结构和分支定界技术结合在一起来求解2DR-SPP,取得了较好的实验结果.Beltrán等人[7]按照随机搜索原则设计的GRASP(greedy randomized adaptive search procedure)算法与VNS(variable neighbourhood search)结合在一起来求解2DR-SPP.而Alvarez-Valdes等人[8]提出了Reactive GRASP算法,该算法分为构建和改进两个阶段,其测试结果优于文献[7].张德富等人[9]则提出了一种有效的砌墙式启发式算法PH,实验结果表明,该算法对规模较大的2DR-SPP能够获得较好的装箱效果.针对2DR-SPP,目前广泛采用的是遗传算法GA(genetic algorithm)、模拟退火算法SAA(simulated annealing algorithm)等搜索算法与BL(bottom-left)、IBL(improved BL)、BLF(bottom-left-fill)等启发式布局算法相结合的方式.Bortfeldt[10]采用了无任何编码的遗传算法来解决矩形装箱问题.Jackobs[11]提出了一种混合算法,通过GA 与BL启发式布局算法相结合,从而将矩形的装箱问题转换成相对简单的装入序列问题.由于BL算法容易出现即使穷举所有情况仍不能得到最优解的现象,因此,Liu[12]提出了一种IBL算法,并与GA相结合,取得了优于文献[11]的装箱效果.Yeung等人[13]针对布料切割问题提出了一种布局算法LFLA,该布局算法的计算复杂度为O(n),相对于BL算法(其复杂度为O(n2)),效率上有了较大的提高.Zhang等人[14]针对矩形装箱问题给出了meta-heuristic算法,该算法主要基于启发式递归策略和SAA.Dereli等人[15]则采用SAA与递归布局方法相结合来解决2DR-SPP.针对不同数量及不同类型的矩形排样,Hopper[16]分别使用了BL,BLF启发式布局算法和GA,NE(naïve evolution),SA,HC(hill-climbing),RS(random search)启发式搜索算法相结合的方式进行求解.迄今为止,虽然对2DR-SPP进行了大量的研究,但从相关文献给出的测试结果来看,该问题仍然有进一步研究的必要.例如,BL,IBL,BLF,LFLA以及贾志欣等人[17]提出的最低水平线LHL(lowest-horizontal-line)等常见的启发式布局算法,它们在对矩形装入的过程中严格按照矩形的某个排列序列进行,容易产生较大的空洞,浪费空间,因此装箱效果不够理想.相对于BL而言,BLF算法可以取得较好的装箱效果,但其算法的时间复杂度却达到了O(n3)[18],不适宜解决规模较大的2DR-SPP.本文针对上述问题设计出一种启发式布局算法,即底部左齐择优匹配算法(lowest-level left align best fit,简称LLABF).本文第1节重点介绍LLABF算法设计.第2节介绍GA+LLABF算法的实现.第3节主要对标准数据集进行模拟测试,并与相关文献结果进行对比分析.第4节给出结论.1 LLABF算法设计1.1 定义设容器宽度所在方向为横坐标X,长度方向为纵坐标Y,容器的左下角为坐标原点(0,0).容器的底部在y=0处,容器可以在Y轴正方向无限延伸.1530 Journal of Software软件学报 V ol.20, No.6, June 2009队列I={π1,π2,…,πn}表示n个矩形的某个装入序列,其中,πi是矩形编号,πi∈[1,n],对任意i,j∈[1,n],当i≠j时, πi≠πj.设I(i)表示队列I中第i个位置上对应的矩形,它由五元组构成,即I(i)={x,y,w,h,θ},其中,I(i).x,I(i).y分别表示该矩形装入后,其左下角的横、纵坐标;I(i).w,I(i).h,I(i).θ分别表示该矩形的宽度、高度和旋转角度.设队列E={e1,e2,…,e m}表示装入过程中产生的轮廓线集合,元素e k为水平线线段(与X坐标轴平行),它由三元组构成,即e k={x,y,w},其中,e k.x,e k.y表示第k个水平线线段的左端点坐标(起点坐标);e k.w表示第k个水平线线段的宽度;并且对任意0<k<m,有e k.x<e k+1.x,即按轮廓线起点的x坐标从小到大排列.队列E具有以下特征:其y坐标具备唯一性,如果相邻线段具有相同的高度y,则进行合并;所有线段在X坐标上的投影不重叠;所有线段的宽度之和刚好等于矩形容器宽度W.最低水平线定义为队列E中其y坐标最小的水平线.最优高度H opt表示穷举所有可能情况得到的最小装箱高度,也称为最优解.装箱高度H packing表示根据当前装入序列,按照布局算法将所有矩形装入矩形容器C内所得到的最小高度.空洞是指在装箱过程中,由矩形或者矩形与容器边界(如x=0,y=0,x=W或者y=H packing)围成的空白区.空洞越多,材料浪费的程度就越严重.1.2 LLABF启发规则设计1.2.1 基本思想一般情况下,对NP难问题很难直接构造出一个最优解或满意解,所以只能通过搜索的方法在整个解空间中寻找最优解或者满意解.当问题规模较大时,这种盲目搜索或者遍历就变得十分困难,甚至根本不可行.因此,相关学者常利用问题解本身的某些结构特征来构造出一些启发式规则,并按照该规则设计出启发式算法,由该算法求得问题的一个满意解.对于矩形条带装箱问题,它有以下几个可以利用的结构特征:(1) 面积较大的矩形装入后产生的空洞较大,面积较小的矩形装入后产生的空洞较小;(2) 面积较大的矩形装入后产生的空洞常常可以装入面积较小的矩形;(3) 装入过程中产生的轮廓线越规整,即组成轮廓线的水平线数量越少,就越有利于后期矩形的装入.本文中,LLABF启发式算法利用了矩形条带装箱问题的3个结构特征,采用了动态选择方法.该方法遵照最佳匹配优先原则来动态选择下一个待装矩形,使得在矩形装入的每一步都尽可能地获得一个较优的装箱结果.重复执行,直至最终获得整体较优解.最佳匹配优先原则(best-fit priority,简称BFP)即在矩形的装入过程中,优先考虑与当前可装入区域宽度或高度相匹配以及可装入的未排矩形,它包括完全匹配优先、宽度匹配优先、高度匹配优先、组合宽度匹配优先以及可装入优先这5种启发式规则.5种启发式规则中完全匹配优先、宽度匹配优先以及组合宽度匹配优先可以减少空洞的产生,特别是在装入初期,由于用来与可装区域进行比较的矩形较多,因此,其匹配的概率越高,就越不容易产生空洞;完全匹配优先以及高度匹配优先使得装箱轮廓相对规整;高度匹配优先以及可装入优先可以将较小矩形延后装入.5种启发式规则结合起来不但可以使矩形装入后产生的空洞数量较少,而且所有的空洞总面积相对也较小.1.2.2 启发规则设计完全匹配优先(full-fit first,简称FFF). 在可装入的轮廓线中选取最低的水平线e k,如果有多个线段,则优先选取最左边的一段.从待装矩形中按照装入序列依次将矩形与e k进行比较,如果存在宽度或者高度与该线段宽度e k.w相等且装入后刚好左填平或者右填平的矩形则优先装入.完全匹配优先能够减少装入后产生的轮廓线数量,使得装入轮廓朝着顶部平齐的方向发展.宽度匹配优先(width-fit first,简称WFF). 在装入过程中,优先装入宽度或者高度与最低水平线e k等宽的矩形,如果存在多个匹配矩形,则优先装入面积最大的.与完全匹配优先规则不同的是,宽度匹配优先并不要求装入后能够实现左填平或者右填平;同时,该规则使得较小矩形有推迟装入的趋势.另外,WFF不会增加装入轮廓线数量.蒋兴波 等:二维矩形条带装箱问题的底部左齐择优匹配算法1531高度匹配优先(height -fit first ,简称HFF ). 在待装矩形中,按照装入序列查询宽度或高度不大于最低水平线e k 宽度且装入后能够实现左填平的矩形,若存在则装入查询到的首个矩形.与FFF 和WFF 不同,HFF 可能会在最低水平线上产生新的、更小的可装入区域,但却增加了轮廓线e k −1的宽度.组合宽度匹配优先(joint -width -fit first ,简称JWFF ). 按装入序列对两个矩形进行组合,如果组合后的宽度与最低水平线宽度e k 相等,则优先装入组合序列中的首个矩形.例如,存在两种组合I (i 1).w +I (j 1).w =e k .w , I (i 2).w +I (j 2).w =e k .w ,如果I (i 1)的面积大于I (i 2),则首先装入I (i 1),否则装入I (i 2).JWFF,FFF 与WFF 规则避免了在最低水平线e k 上产生新的、更小的可装入区域,从而减少了整个装箱过程产生空洞的可能性.为了保证时效性,算法中设置了查询范围searchNum ,即从当前位置开始,最多可以进行searchNum ×searchNum 次连续矩形的两两组合查询.可装入优先(placeable first ,简称PF ). 在一定范围内,从待装矩形件中按照装入序列依次查找宽度或高度不大于最低水平线e k 宽度的矩形,若存在,则将其装入;若存在多个,则装入面积最大的矩形.PF 可能在最低水平线上产生新的、更小的可装入区域,同时使得较小矩形延迟装入.在矩形的装入过程中,满足FFF,WFF,HFF 以及JWFF 规则的矩形可能并不存在,此时就必须考虑PF.如果满足PF 规则的矩形仍不存在,那么必然会在最低水平线上产生空洞.与BL,IBL,LFLA 以及LHL 算法不同,该空洞是在一定范围内由于不存在可装入的矩形而产生的,因此其面积更小.与BLF 算法不同,LLABF 算法并不需要保存新产生的空洞.图1给出了按照最佳匹配优先原则的部分启发规则实现矩形装入的示意图,其中,矩形的装入序列是4 2 7 6 5 1 8 3.图1(a)表示已经装入了4个矩形,产生了由4个线段e 1,e 2,e 3,e 4组成的装入轮廓,其中,e 2为最低水平线,其对应的区域是接下来首先要考虑的区域.图1(b)采用完全匹配优先:与最低水平线e 2等宽的矩形有1号和3号,但由于只有3号矩形装入后能够实现左填平,因此首先选择该矩形装入.图1(c)采用宽度匹配优先:满足该条件的矩形有1号和3号,但由于1号矩形面积较大,所以优先装入.图1(d)采用高度优先:1,8,3这3个矩形都满足其宽度不大于e 2.w ,但只有8号和3号矩形装入后能够实现左填平,且8号矩形位置在3号之前,所以优先装入.(a) (b)(c) (d)Fig.1 Packing result according to BFP principle图1 按BFP 原则的装箱结果图1.3 LLABF 算法的实现LLABF 算法是按照BFP 原则进行装箱的启发式布局算法,其算法实现如下所示.1532 Journal of Software软件学报 V ol.20, No.6, June 2009LLABF (input: I; output: H packing)beginInitContours(E); InitCoordinate(I); H packing=0; // 初始化;H packing:装箱高度① for (ii=1; ii≤n; ii++)begin②if (I(ii).x>−1) goto ①; // 当前矩形件已被装入在E中取最低水平线e k;j=−1; //记录用于装入的矩形件所在序列号FullFitFirst(ii,I,e k,j); if (j>−1) goto ④;//完全匹配FFFWidthFitFirst(ii,I,e k,j); if (j>−1) goto ④; //宽度匹配WFFHeightFitFirst(ii,I,e k,j); if (j>−1) goto ④; //高度匹配HFFJointWidthFitFirst(ii,I,e k,7,j); if (j>−1) goto ④; // 组合宽度匹配JWFFPlaceableFirst(ii,I,e k,n/6,j); if (j>−1) goto ④; //可装入优先PF③合并E中的边,转②;④将I集合中第j位元素对应的矩形件I(j)装入在E中的最低水平线e k上;如果未旋转,则H packing=max(H packing,I(j).x+I(j).h),否则,H packing=max(H packing,I(j).x+I(j).w);更新集合I和E;如果j>ii,则转②;endend其中,InitCoordinate(I)是对所有矩形块的坐标x,y进行初始化.在初始阶段,由于所有的矩形都未进行装入, InitCoordinate(I)将其坐标x,y全部初始化为−1,即对任意i∈[1,n],I(i).x=I(i).y=−1.InitContours(E)对装箱轮廓线进行初始化.在初始阶段,由于容器C未被装入,因此,其轮廓线仅由一个水平线线段构成,即E={e1},其中,e1={0,0,W},W为容器C的宽度.FullFitFirst,WidthFitFirst,HeightFitFirst,JointWidthFitFirst,PlaceableFirst分别表示按照FFF,WFF,HFF, JWFF以及PF等启发式规则在队列I中第ii位开始向后依次查找与最低水平线e k相匹配矩形的查询过程,其返回值为j.其中,JointWidthFitFirst,PlaceableFirst中的查询范围分别设置为7,n/6,为多次实验后取得的经验值.1.4 LLABF算法举例图2(a)是5个矩形构成的一个完美装箱图,其装箱高度H packing等于最优高度H opt,空间的利用率为100%.对于BL,IBL,LFLA以及LHL算法,要实现图2(a)的完美装箱(不考虑旋转),其装入序列就必须为1 2 3 4 5.对于序列为1 3 4 5 2和1 3 2 4 5,其装箱结果分别如图2 (b)和图2(c)所示,这两种装入序列都会产生较大的空洞.(a)(b) (c)Fig.2 Packing result according to BL, IBL, LFLA or LHL algorithm图2 BL,IBL,LFLA或LHL算法的装箱结果图对于同样的装入序列,采用LLABF算法则能得到如图2(a)所示的完美装箱.图3显示了装入序列为1 3 4 5蒋兴波 等:二维矩形条带装箱问题的底部左齐择优匹配算法15332、采用LLABF 算法得到的装箱结果(其中,*表示该位置上的矩形已经装入容器中).初始化:E ={e 1},e 1={0,0,W };对任意i ∈[1,n ],I (i ).x =I (i ).y =−1; 开始阶段虽然所有矩形都不满足完全匹配FFF 、宽度匹配WFF 以及高度匹配,但I (1).w +I (5).w =e 1.w ,满足两两组合宽度匹配JWFF,所以将编号为1的矩形装入在矩形容器C 的左下角;装入结果如图3(c)所示,此时,E ={e 1,e 2};取最低水平线e 2,按照LLABF 启发式规则依次查询匹配矩形,其中I (5),即编号为2的矩形满足WFF,故将其装入在e 2上;装入结果如图3(d)所示,此时,E ={e 1,e 2};以此类推,其装入过程如图3(e)~图3(g)所示.根据LLABF 算法,如果装入序列符合1 2 3 * *或者1 3 * * * 模式(其中,*代表未列出的矩形编号),其结果都与1 2 3 4 5序列的装箱结果相同,即都能得到如图3(a)所示的完美装箱.在图3中,满足第1种模式的装入序列包括1 2 3 4 5,1 2 3 5 4两种,满足第2种模式的装入序列包括1 3 2 4 5,1 3 4 2 5,1 3 2 5 4,1 3 5 2 4,1 3 5 4 2,1 3 4 5 2这6种.针对图2中的5个矩形的装箱问题,LLABF 找到最优解的概率是BL,IBL,LFLA 以及LHL 的8倍之多.因此,相对于传统布局算法,LLABF 算法更容易找到最优解.Fig.3 Packing result according to LLABF algorithm图3 LLABF 算法的装箱结果图2 GA +LLABF 算法实现2.1 算法的总体流程 本文采用遗传算法GA 与LLABF 相结合的方式来解决二维矩形条带装箱问题.其中,GA 用来实现矩形装入序列的初步确定,LLABF 则根据自身的启发式规则将矩形装入到矩形容器中.GA+LLABF 的算法流程如下所示.(1) 个体染色体编码.(2) k =0;随机产生m 个个体组成初始群体pop (k )={I 1,I 2,…,I m }.(3) 对单个个体I i 执行LLABF(Ii ,H packing );记录个体适应度f (I i )(f (I i )=1/H packing ).(4) 判断是否满足停止条件.若是,则停止计算,输出最佳结果;否则,继续.(5) 利用排序选择操作选择m 个个体形成新的群体selpop (k +1).(6) 根据交叉概率P c 进行交叉操作,产生群体crosspop (k +1).(a) (b) 1 3 4 5 2(c) * 3 4 5 2 (d) * 3 4 5 * (e) * * 4 5 * (f) * * * 5 * (g) * * * * *1534 Journal of Software软件学报 V ol.20, No.6, June 2009(7) 根据变异概率P m进行变异操作,产生群体mutpop(k+1).(8) pop(k+1)=mutpop(k+1);k=k+1;转(3),循环.2.2 编码及适应度函数矩形装箱采用整数编码(encoding),即n个矩形分别用整数1,2,…,n编号,矩形的一个装箱方案对应于一个染色体编码,I j=(π1,π2,…,πn),其中,|πi|表示矩形编号(1≤|πi|≤n),当πi<0时,矩形旋转90°,否则不旋转;I j表示第j个个体.例如:个体I j的染色体编码为(−5 3 −7 6 1 10 8 −2 4 −9),表示装入序列为5 3 7 6 1 10 8 2 4 9.在装入过程中,编号为5,7,2,9的矩形进行90°旋转.适应度函数(fitness function)可以由装箱高度的倒数表示,即f(I j)=1/H packing(I j).2.3 选择为了避免丢失上一代产生的最佳个体,本文在采用排序选择方法(rank-based select model)的同时,还混合使用了最优保存策略(elitist model).排序选择方法主要依据个体适应度值之间的大小关系,对个体适应度是否取正值或负值以及个体适应度之间的数值差异程度并无特别要求.其步骤如下:(1) 对群体中所有个体按照适应度从大到小排序.(2) 根据下列公式计算每个个体的选择概率:()iP=α×(1−α/m)I,sP表示排列在第i个位置的个体的选择概率,m表示群体大小,α为调节参数.其中,()is(3) 根据步骤(2)计算得出的个体概率值作为其能够被遗传到下一代的概率,基于该值应用比例选择(赌盘选择)的方法产生下一代群体.2.4 交叉(crossover)交叉是产生新个体的主要方法.本文主要采用环形部分交叉(circular-based part crossover):①随机生成起始交叉点P cr∈[1,n];②随机生成交叉长度L cr∈[1,n];③将P cr后的L cr位基因进行交换,如果P cr+L cr>n,则将染色体前P cr+L cr−n位基因进行互换;④然后依次将未出现的基因填入空白基因座.2.5 变异(mutation)变异是产生新个体的辅助方法,它决定了遗传算法的局部搜索能力,保持群体的多样性.本文中的变异包括两部分:一是旋转变异,二是装入序列变异.(1) 旋转变异旋转变异采用了单点取反和环形部分取反两种方式.单点取反(single point reversion)变异:①随机生成变异点P mu∈[1,n];②取反.例如,对于个体I j=(π1,π2,…, πi,…,πn),P mu=i,则变异后产生的新个体为I′=(π1,π2,…,−πi,…,πn).j环形部分取反(circular-based part reversion)变异:①随机生成变异点P mu∈[1,n];②随机生成变异长度L mu∈[1,n];③将P mu后的L mu位取反,如果P mu+L mu>n,则将染色体首部的P mu+L mu−n位基因取反.(2) 装入序列变异装入序列变异采用了两点位置互换与环形部分逆转两种方式.两点位置互换(two point exchange)变异:①随机生成两个变异点i,j∈[1,n];②将两个基因互换.例如:I j=(π1, π2,…,πi1,…,πi2,…,πn),i1,i2∈[1,n],则变异后产生的新个体为I′=(π1,π2,…,πi2,…,πi1,…,πn).j环形部分逆转(circular-based part inversion)变异:①随机生成变异点P mu∈[1,n];②随机生成变异长度L mu∈[1,n];③将P mu后的L mu位逆转,即将(P mu+i) mod n(即P mu+i对n取模)与(P mu+L mu−i) mod n互换,i∈[0,L mu/2].蒋兴波等:二维矩形条带装箱问题的底部左齐择优匹配算法15353 实验结果实验平台:Pentium 4 3.0GHz,512M,Windows Server 2003.运行参数:交叉概率P c=0.95,变异概率P m=0.85.本文的实验针对两组标准测试问题来进行,第1组来自文献[11],第2组来自文献[16].3.1 实验1文献[11]给出了两组标准测试实例,均由40×15的大矩形切割而成.其中,第1组实例由25个矩形组成,第2组实例由50个矩形组成.本文实验中,GA的群体大小和运行代数参照文献[12],分别设为20和100.对每组实例皆重复运行100次.实验结果对比见表1.Table 1Computational results of four algorithms表14种算法的计算结果Jakobs S[11] GA+BL Liu DQ[12]GA+IBLBortfeldt A[10]SPGALGA+LLABFW×H nH min H avg H min H avg H min H avg H min H avg H max40×15 25 17 17.481616.9716 16.001515.001540×15 50 17 17.281617.0115 15.001515.0015 表1中,H min,H max,H avg分别指多次运行中得到的最小H packing、最大H packing及平均H packing.由表1可以看出,GA+LLABF算法在对2组标准实例进行的100次测试中均找到了最优解15,而其他3种算法中,除了SPGAL找到了第2组数据集的最优解之外,其余算法求得的H packing均大于H opt.因此,实验结果表明,本文的GA+LLABF算法更加有效.3.2 实验2在文献[16]中,Hopper等人给出了21组不同尺寸的测试实例,这些实例分成7个大类,每个大类由3组实例组成,其矩形个数从16个~197个不等.每个实例的最优高度H opt已经给出,其相关数据见文献[16].在本文的算法中,初始种群大小参照文献[16],设为50,终止条件为进化代数等于 2 500代,或者装箱高度H packin等于最优解H opt.对每组测试实例重复运行10次,记录每个大类30次(每个大类3个实例,每个实例进行10次测试)测试的平均装箱高度到最优解的相对距离RDBSOH,RDBSOH(%)=100×(30次平均H packing−H opt)/H opt.实验结果对比见表2.Table 2Computational results of 12 algorithms (relative distance of best solution to optimum height (%))表2 12种算法的计算结果(RDBSOH(%))Algorithm C1 C2 C3 C4 C5 C6 C7 AverageGA+BLF[16] 4 7 5 3 4 4 5 4.57 SA+BLF[16] 4 6 5 3 3 3 4 4.00 GRASP+VNS[7]14.40 17.33 12.93 6.80 4.51 3.55 2.89 8.92 HR[3]8.33 4.45 6.67 2.22 1.85 2.5 1.8 3.97 Hybrid SA[15] 1.66 4.88 4.00 3.94 3.18 3.30 3.38 3.48SA+HR[14] 5.00 4.47 2.23 2.22 1.86 2.5 3.24 3.07 PH[9] 5.00 4.44 4.44 3.33 1.11 1.11 1.25 2.95HRBB[6] 1.70 0.00 1.10 2.20 1.90 1.40 1.30 1.40GRASP[8]0.00 0.00 1.08 1.64 1.10 1.56 1.36 1.33SPGAL[10] 1.70 0.90 2.20 1.40 0.00 0.70 0.50 1.00 A1[5]0.00 0.00 1.11 1.67 1.11 0.83 0.42 0.73 GA+LLABF 0.00 0.00 0.00 0.00 0.00 0.00 0.40 0.06 从表2可以看出,在现有的算法中,按照Averag e即平均RDBSOH,位居前4位的较优算法依次为A1, SPGAL,GRASP与HRBB,它们对7大类实例计算得出的Average分别为0.73,1.0,1.33和1.4.而本文给出的GA+LLABF算法计算得出的Average为0.06,其中6大类实例的RDBSOH为0,即对应的所有实例其10次实验结果皆取得了最优解.因此,实验结果进一步表明,针对2DR-SPP,本文提出的GA+LLABF算法优于其他算法.1536Journal of Software 软件学报 V ol.20, No.6, June 2009表3给出了GRASP+VNS,GRASP,HRBB 以及本文的GA+LLABF 对21组实例求得的实际测试结果H packing .在GA+LLABF 算法对21组实例的测试中,10次实验皆能获得最优解H opt 的有18组(21组中,除C 71,C 72,C 73外),占全部实例的86%(18/21),而GRASP 为38%(8/21)、HRBB 为33%(7/21);10次实验中至少获得1次最优解的实例有19组(即最小高度H min =最优高度H opt ),占全部实例的91%(19/21),而GRASP 为38%(8/21)、HRBB 为43%(9/21);10次实验获得的最大高度H max 与最优解H opt 之间的距离最多为1个基本单位.因此,无论是最优解所占比例还是最小高度H min 、平均高度H avg 以及最大高度H max ,GA+LLABF 求得的结果均具有明显的优势.Table 3 Computational results (H packing ) of 21 instances表3 21组实例计算结果 Beltrán DJ GRASP+VNS [7] (20 runs) Alvarez-Valdes R GRASP [8] (10 runs) Cui YD HRBB [6] (6 runs) GA+LLABF(20 runs) CategoryH opt n H min H avg H max H min H avg H min H avg H max H min H avg H max C 1120 16 21 22.30 23 20 20.00 2020.0020 20 20.00 20 C 1220 17 21 23.20 25 20 20.00 2121.0021 20 20.00 20 C 1320 16 22 23.15 24 20 20.00 2020.0020 20 20.00 20 C 2115 25 16 17.20 19 15 15.00 1515.0015 15 15.00 15 C 2215 25 16 18.00 20 15 15.00 1515.0015 15 15.00 15 C 2315 25 16 17.60 20 15 15.00 1515.0015 15 15.00 15 C 3130 28 32 32.55 34 30 30.00 3030.0030 30 30.00 30 C 3230 29 32 34.25 37 31 31.00 3131.0031 30 30.00 30 C 3330 28 33 34.85 37 30 30.00 3030.0030 30 30.00 30 C 4160 49 63 63.55 66 61 61.00 6060.5061 60 60.00 60 C 4260 49 61 64.60 68 61 61.00 6161.3362 60 60.00 60 C 4360 49 62 64.10 66 61 61.00 6161.0061 60 60.00 60 C 5190 73 92 93.45 95 91 91.00 9090.6791 90 90.00 90 C 5290 73 93 94.75 97 91 91.00 9292.0092 90 90.00 90 C 5390 73 92 94.00 98 91 91.00 9191.1792 90 90.00 90 C 61120 97 123 124.40 130121121.90121121.00121120 120.00 120C 62120 97 123 124.65 128121121.90121121.83122120 120.00 120C 63120 97 122 123.75 126121121.90121121.33122120 120.00 120C 71240 196 244 246.45 248244244.00242242.17243241 241.00 241C 72240 197 244 247.05 255242242.90245245.00245241 241.00 241C 73 240 196 245 247.30 258243243.00241241.33242240 240.90 241图4(H packing =H opt =120)、图5(H packing =H opt =240)显示了由GA+LLABF 算法求得的部分实例装箱图,其装箱高度皆为最优高度.Fig.4 Packing result of C 63 Fig.5 Packing result of C 73图4 C 63的装箱结果图 图5 C 73的装箱结果图 LLABF 算法在对矩形的装入过程中,每一步均按照FFF,WFF,HFF,JWFF 以及PF 规则依次查询与最低水平蒋兴波等:二维矩形条带装箱问题的底部左齐择优匹配算法1537线相匹配的矩形.对于FFF,如果某个阶段存在i个待装矩形,那么最坏情况下将比较i次;其余4个启发式规则除了JWFF的比较次数为常数K以外,皆与完全匹配优先类似,因此,单步比较次数最坏情况下为4i+K.将所有n 个矩形全部装入容器中,总的比较次数最坏情况下为4×(n+n−1+n−2+…+2+1)+n×K,共2n×(n+1)+n×K,因此其复杂度为O(n2).对于矩形在容器中的装入,主要的比较次数在于查找最低水平线,而最低水平线的数量不会超过n,所以,其复杂度最坏情况下为O(n2).整体上,LLABF的时间复杂度为O(n2).同时,对于GA,由于群体规模和进化代数皆为常数,其本身的比较次数也为常数,所以GA+LLABF算法的复杂度主要由LLABF决定,即最坏情况下为O(n2).表4给出了表2中所有算法的计算时间.由于算法的实现方式、测试平台与环境的不同,要对算法的快慢进行精确比较是不科学也是不可行的.但是从表4中,我们能够大致看出各种算法的运行快慢程度.由于本文算法采用了装箱高度等于最优解作为终止运行条件之一,因此,虽然C5的规模大于C4,但其获得最优解的速度明显快于C4,所以运行时间相对较小.Table 4Run times of 12 algorithms (s)表412种算法的运行时间(s)Algorithm C1 C2 C3 C4 C5 C6 C7 Averagea[16]a SA+BLF[16]42.00 144.00 240.00 1 980.00 6 900.0022 920.00250 860.00 40 441.86b GRASP+VNS[7] 0.00 0.00 0.00 0.00 0.02 0.07 1.37 0.21c HR[3]0.000.000.03 0.14 0.69 2.21 36.07 5.59dHybrid SA[15] 3.64 6.36 18.78101.20287.27 757.20 1650.6 403.57e SA+HR[14] 23.20 42.80 59.20170.00343.60 444.90 1328.80 344.64f PH[9]0.00 0.00 0.00 1.51 5.69 23.74 707.12 105.45g HRBB[6] 0.27 0.33 0.88 2.19 2.83 2.24 4.26 1.86h GRASP[8]- - -- ---60.00i SPGAL[10]--- ----139.00j A1[5]0.370.611.71 0.15 0.40 3.98 45.02 7.46GA+LLABF0.000.032.233.6 1.3 33.2 328.3 52.67a—Pentium Pro 200MHz and 65MB memory c—Dell GX260 with a 2.4GHz CPUe—Dell GX270 with a 3.0GHz CPU g—Pentium 4 2.8GHZ, 512MB memory i—Pentium PC with 2GHz b—Pentium 166MHz and 96MB RAM d—Pentium Ⅲ 797MHZf—Pentium 4 1.6GHZ, 256MB memory h—Pentium 4 mobile 2.0GHZ, 512MB memory j—2.4GHz PC with 512MB memory4 结论与展望本文分析了传统布局算法的不足,提出了一种启发式布局算法LLABF,并将其与GA相结合.实验结果表明,GA+LLABF算法解决2DR-SPP问题更加有效.但是要评价一种算法的优劣,除了要考虑算法的求解结果以外,还应兼顾算法的运行效率.因此,改进算法以进一步提高运算速度,将是我们下一步的研究重点.References:[1] Hifi M. Exact algorithms for the guillotine strip cutting/packing problem. Computers & Operations Research, 1998,25(11):925−940.[2] Lesh N, Marks J, McMahon A, Mitzenmacher M. Exhaustive approaches to 2D rectanglar perfect packings. Information ProcessingLetters, 2004,90:7−14.[3] Zhang DF, Kang Y, Deng AS. A new heuristic recursive algorithm for the strip rectangular packing problem. Computers &Operations Research, 2006,33(8):2209−2217.[4] Chen DB, Huang WQ. Greedy algorithm for recatngle-packing problem. Computer Engineering, 2007,33(4):160−162 (in Chinesewith English abstract).[5] Chen M, Huang WQ. A two-level search algorithm for 2D rectangular packing problem. Computers & Industrial Engineering, 2007,53(1):123−136.。
工业机器人外文参考文献
工业机器人外文参考文献工业机器人外文参考文献1. Gao, Y., & Chen, J. (2018). Review of industrial robots for advanced manufacturing systems. Journal of Manufacturing Systems, 48, 144-156.2. Karaman, M., & Frazzoli, E. (2011). Sampling-based algorithms for optimal motion planning. The International Journal of Robotics Research, 30(7), 846-894.3. Khatib, O. (1986). Real-time obstacle avoidance for manipulators and mobile robots. The International Journal of Robotics Research, 5(1), 90-98.4. Lee, J. H., & Kim, E. B. (2019). Energy-efficient path planning for industrial robots considering joint dynamics. Journal of Mechanical Science and Technology, 33(3),1343-1351.5. Li, D., Li, Y., Wang, X., & Li, H. (2019). Design and implementation of a humanoid robot with flexible manipulators. Robotica, 37(12), 2008-2025.6. Liu, Y., Ren, X., & Wang, T. (2019). A novel adaptive fuzzy sliding mode control for a nonholonomic mobile robot. Soft Computing, 23(3), 1197-1209.7. Loh, A. P., & Tan, K. K. (2014). Vision-based controlof industrial robots using an adaptive neural network. Robotics and Computer-Integrated Manufacturing, 30(2), 177-184.8. Niemeyer, G., & Slotine, J. J. (1990). Stable adaptive teleoperation. The International Journal of Robotics Research, 9(1), 85-98.9. Park, J. Y., & Cho, C. H. (2018). Optimal path planning for industrial robots using a hybrid genetic algorithm. International Journal of Precision Engineering and Manufacturing, 19(5), 755-761.10. Wang, Z., Wang, X., & Liu, Y. (2019). Design and analysis of a novel humanoid robot with intelligent control. International Journal of Advanced Manufacturing Technology, 102(5-8), 1391-1403.。
遗传禁忌搜索算法收敛性和时间复杂度分析
遗传禁忌搜索算法收敛性和时间复杂度分析牟乃夏;徐玉静;李洁;张灵先【摘要】遗传禁忌搜索算法多用于车辆路径优化、旅行商问题等,试验证明:融合遗传算法与禁忌搜索算法的混合算法相比单一算法的性能有较大提升,但缺少理论证明.本文阐述了遗传禁忌搜索算法的混合策略,从理论上对该算法的收敛性进行了证明,对时间复杂度进行了分析.应用马尔科夫链模型证明了遗传禁忌搜索算法是以概率1收敛到全局最优解的,并应用求解随机算法时间复杂度的方法,即求解算法的期望收敛时间,估算了该算法的时间复杂度,结果证明该算法的时间复杂度与所得解的多样性、问题规模以及遗传算法的种群数量有关.%The hybrid of genetic algorithm and tabu search algorithm has greatly improved performances relative to one single algorithm.It has been widely used in vehicle route optimization and travel planning,etc.In this paper,a hybrid strategy of genetic tabu search algorithm was introduced and its convergence was theoretical proved.The time complexity of the algorithm was analyzed as well.By using Markov chain model,the algorithm of genetic tabu search was proved to have the global optimal solution with converge of probability 1.Mean-while,its time complexity was calculated by a stochastic algorithm.Based on the above methods,the time com-plexity of the algorithm was obtained,which was proved to be highly related to the diversity of the solution,the problem scale and the population of the genetic algorithm.【期刊名称】《河南理工大学学报(自然科学版)》【年(卷),期】2018(037)004【总页数】5页(P118-122)【关键词】遗传算法;禁忌搜索算法;收敛性;时间复杂度;马尔科夫链模型【作者】牟乃夏;徐玉静;李洁;张灵先【作者单位】山东科技大学测绘科学与工程学院,山东青岛266590;中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室,北京100101;山东科技大学测绘科学与工程学院,山东青岛266590;山东科技大学测绘科学与工程学院,山东青岛266590;山东科技大学测绘科学与工程学院,山东青岛266590【正文语种】中文【中图分类】P2080 引言遗传算法(genetic algorithm,GA)是由美国Michigan大学的Holland教授首先提出的[1],它是通过模拟自然界生物进化的过程与机制来求解极值问题的一类自组织、自适应人工智能的技术[2]。
模具英文词汇
注册登录论坛开通窝窝网页游戏打折机票搜索帮助导航繁體中文中华模具精英社区»出口交流» [分享] 模具英文词汇对照与塑料模专业英词窗体顶端窗体底端返回列表回复发帖狂送千万好礼年终巨献让利空前发新话题发布投票发布悬赏发布辩论发布活动发布视频发布商品发短消息加为好友1# 打印hai 当前离线UID9帖子439精华6积分596威望81金钱184阅读权限50在线时间40 小时注册时间2009-8-6最后登录2010-10-11高级设计师UID9帖子439精华6积分596威望81金钱184阅读权限字体大小: tThai发表于2009-8-6 14:27 | 只看该作者踩窝窝送礼物问候Ta[分享] 模具英文词汇对照与塑料模专业英词窗体顶端窗体底端关于“模流分析相关...”的内容本站搜索更多关于“模流分析相关英文词汇”的内容[分享] 模具英文词汇对照与塑料模专业英词很多初学者找不到相关对照表,我现在发上来供大家学习,如果还有没有说到的,请大家补上来,英文詞匯一、模具結構Mould / Tooling --- 模具Top Clamp Plate ---面板Mould Base ------模胚Runner Stripper Plate---- 水口板Gate ----------------澆口(入水位)A Plate ------------- A 板Side Gate ----------側澆口Stripper Plate------推板Pin Point Gate-----點澆口B Plate--------------B 板Submarine Gate---潛伏澆口Support Plate------托板Disk Gate----------------- 圓盤澆口Spacer Block------------- 方鐵Tab Gate------------------ 薄片澆口Ejector Retainer Plate--- 面針板Spoke Gate--------------- 輪輻澆口Ejector Plate------------- 底針板Direct Gate--------------- 直接澆口Bottom Clamp Plate---- 底板Cavity--------------------- 型腔Insert---------------------- 鑲件Core----------------------- 型芯50在线时间40 小时注册时间2009-8-6 最后登录2010-10-11 二、加工工具Bushing Borers----- 杯司鑽頭Tapered Reamers--- 爹把絞刀Core Pins------------ 哥針Ejector Pin Reamers----- 頂針絞刀Capscrew Counterbores--杯頭螺絲撥頭Leader Pin Reamers----- 導邊絞刀Ejector Pin Counterbores----- 頂針撥頭Springs“A”Type (Orange Color)------扁線壓力彈弓“A”(橙色) Leader Pin Counterbores----- 導邊撥頭Springs“B”Type (Silver Color)------- ----扁線壓力彈弓“B”(銀色) Return Pin Counterbores----- 回針撥頭Tapered Drills----------------- 爹把鑽Shoulder Bushing Counterbores-------- 有托杯司撥頭“R”Cutters-----------------唧咀弧度刃三、螺紋標準名稱BSW----------------- 英制粗螺紋Metric (Coarse)---- 公制粗螺紋BSF ----------------- 英制幼螺紋Metric (Fine)------- 公制幼螺紋BSPP---------------- 英制管平衡螺紋Pg------------------- 公制電力螺紋BSPT---------------- 英制管斜度螺紋UNC---------------- 美制粗螺紋BA------------------ 英國協會螺紋UNF---------------- 美制幼螺紋四、標準模具零件Angular Pins------- 斜度邊釘Return Pins--------- 回針Dowel Pins--------- 管位釘Shoulder Bushings--有托杯司E.G. Bushing--------中托司套Shoulder Leader Pins--有托導邊Ejector Pin---------- 圓身頂針Special Water Junctions--特殊喉咀Hotwork Ejector Pins--- 耐熱頂針Spring-------------- 彈簧HS Ejector Pins--- 高速度鋼頂針Sprue Bushings--- 唧咀Ejector Blade----- 扁身頂針Stop Pins ---------- 限位釘Ejector Return Assembly--- 頂針板回歸器Straight Bushings-------- 直身杯司Ejector Sleeves----------- 頂針司筒Straight Leader Pins------直身導邊Gas Expeller------------------- 疏氣針Support Pillars ---------------- 撐頭Jiggle pin---------------------- 斜頂Support Pins------------------ 細水口導邊Locating Rings--------------- 定位圈Tapered Interlocks----------- 爹把邊釘“P”Locks------------------ 定位扣Tension Link-------------------- 拉力板Puller Bolt Collars------------- 拉桿介指Thermo Pin--------------------- 導熱針Puller Bolts--------------------- 脫水口螺絲Water Junctions---------------- 標准喉咀Puller Lock----------------------拉模扣Hot Runner System------------ 熱流道系統五、不良成形現象Air Trap----------------------- 困氣Flash--------------------------- 披鋒Air Vent----------------------- 排氣Jetting------------------------- 摺紋Black Spots------------------- 黑點Short Shots-------------------- 走膠不全Burning 燒焦Sinks------------------------- 局部收縮Burr--------------------------- 毛刺Splay-Silver Streaks-------- 水花白Cold Well--------------------- 冷料穴Weld Line--------------------- 夾水紋Distortion--------------------- 扭曲六、圖紙用字Drawn By------------- 繪圖者Chamfer-------------- 倒角Checked By---------- 核對者Fillet------------------ 倒圓角Approved by--------- 批核者Draft Angle---------- 脫模斜度Client / Customer--- 客戶Parting Line---------- 分型線Title ------------------- 題目Center Line---------- 中心線Drawing No---------- 圖紙編號Tolerance ------------- 公差Part name------------ 零件名稱Detail View---------- 詳細圖Tool No--------------- 模具編號Top View ------------- 頂視圖Tool Type------------ 模具種類Front View----------- 正視圖No of Cavity--------- 模腔數Side View------------ 側視圖Shrinkage------------------ 收縮率Section View--------------- 切面圖Issue------------------------ 發行Notes------------------------ 告示Modification--------------- 更改Wall Thickness------------ 膠位厚度Item ------------------------ 項目Texture--------------------- 花紋Material-------------------- 材料Radius---------------------- 半徑BOM------------------------ 物料菜單Diameter------------------- 直徑Date ------------------------ 日期Colour---------------------- 顏色Scale------------------------ 比例Coarse---------------------- 粗NTS------------------------- 不按比例Fine------------------------- 幼Unspecified---------------- 未註明的Projection------------------ 投影七、一般加工方法Boring---------------- 鏜加工Milling---------------- 銑床加工CNC------------------ 數控銑床加工Polishing------------- 拋光Sand Blasting-------------- 噴砂Engraving------------------ 雕刻Sand Casting--------------- 砂模鑄造Etching--------------------- 蝕紋Sawing---------------- 鋸Forging--------------- 鍛造Sintering-------------- 燒結Grilling--------------- 鑽床加工Tap-------------------- 攻螺絲Grinding-------------- 磨床加工Turning--------------- 車床加工Heat Treatment------ 熱處理Welding-------------- 燒焊Latheing-------------- 車床加工Wire Cut-------------- 線切割加工八、其他Injection-------------- 注射Key-------------------- 鑰匙Plastic----------------- 塑膠Lens------------------- 鏡片Rubber---------------- 橡膠Handle---------------- 手柄Resin------------------ 樹脂Cover----------------- 蓋Feet------------------- 英尺Panel------------------ 表板Inch------------------- 英吋Shaft------------------ 軸Meter----------------- 米/ 公尺Motor----------------- 馬達Millimeter------------ 毫米Rust------------------- 生鏽Height---------------- 高度Dimension----------- 尺寸Depth----------------- 深度Rib-------------------- 骨位Weight---------------- 重量Glass Fiber----------- 玻璃纖維Speed----------------- 速度Friction--------------- 摩擦力Temperature--------- 溫度Tension--------------- 拉力Screw----------------- 螺絲Pressure-------------- 壓力Screw Driver-------- 螺絲刀Alarm----------------- 警報Thread---------------- 螺紋Pneumatic------------ 氣動Pitch------------------ 螺距Hydraulic------------- 液壓Base------------------- 基座Cylinder-------------- 唧筒Button---------------- 按鈕Lubricating Oil------ 潤滑油Gear------------------- 齒輪Magnet--------------- 磁石Pulley----------------- 滑輪Spanner--------------- 扳手Housing--------------- 外殼Fixture---------------- 夾具收藏分享评分回复引用订阅TOP发短消息加为好友hai 当前离线UID9帖子439精华6积分596威望81金钱184阅读权限50在线时间2#hai发表于2009-8-6 14:28 | 只看该作者踩窝窝送礼物问候Ta常用朮語中英文對照short shot缺膠break斷裂flash披鋒gap too large after assembly裝配間隙大dimension NG尺寸NGfailed to assembked不能裝配no conduct due to broken wire銅絲線斷,不導電力度不良defective forcedistortion變形poor plating電鍍不良damage破損returned by production line生產線退貨wire come off端子松脫printing defect tapping defect打印不良,滑絲40 小时注册时间2009-8-6 最后登录2010-10-11高级设计师UID9帖子439精华6积分596威望81金钱184阅读权限50在线时间40 小时注册时间2009-8-6 最后登录2010-10-11 hurt批傷Crack爆裂bruise碰傷broken snap斷扣Rust生銹Reworked by supplier供應商返工Hurt on wire線披傷Holes有針孔Scratch划傷Foreign material異物Dust灰塵Cable come off線脫落Press hurt壓傷Which caused the abnormal voice from assembly造成裝配有異音Broken pins扞腳斷Lead wire導線Motor馬達It cannot rotate smoothly轉動不暢Loose gear齒輪松Lead wire導線Loose tube and bad insertion wire線頭熱熔膠管脫落Panel按健表板Key defect按健不良Fold皺折Structure different with sample結构與樣板不符Broken tab柱子斷Lead wire assy導線組Adaptor火牛Black specks黑點Hole blocked up塞孔Crack 破裂Mixed color料花Spec標准Actual實測Spring彈弓Non homogeneouse colour混色Hole larger孔徑大There are oil on the part surface部品表面有油Gear dent齒輪位碰傷Thickness NG厚度NGE-PIN too high頂針位高Higher gate水口位高Colour different異色Contamination 雜物Grease 油污Cosmetic defects熔膠不良Rusting 銹斑Resilience NG彈力NGSmudge 污跡Without chamfer未倒角Wire is broken線斷White plastie of terminal peal off白膠脫落Wire is broken線破損Broken snap扣位斷Loose connector線頭松脫焊線脫落The welded wire looseThe boss is shorter柱子低Oddment 碎屑Rand pick 選別游標卡尺/Vemier Caliper ( VC)電子卡尺/Digital Caliper( DC)三座標測量儀(三次元)/Coordinate Measuring Maching(CMM) 電子千分尺/Digital Micrometer( DM)游標千分尺/Vernier Micrometer(VM)針規/Pin Gauge( PG)投影儀/Projector( P)電子高度尺/Digital Height Gauge( DH)游標高度尺/ V emier Height Gauge( VH)塞尺/Thickness Gauge( TG)直尺/Ruler( R)百分表/Dial Indicator( DI)涂層厚度計/Coating Thickness Gauge( CT)塊規/Block Gauge( BG)角度規(尺)/Goniometer/Bevel Protracter( G)直角尺/Right Angle Ruler/Spuare( AR)溫適度計表/Hygrothermograph( H)平台板/Flat/Plate( F)色差儀/Chromometer( C)鈕具表/Torfque Gauge( TE)推拉力計/Pull-Push Gauge( PP)數字鈕鉅儀/Digital Torfque Gauge( DT)電子稱/ Digital Scale( DS)法碼/Weights( WT)天平/Balance( BG)照度計/Light Meter( LM)萬用表/Multimeter(M)螺紋規/Thread Gauge( TD) 表面粗糙度樣塊/Roughness Block( RB) 功力表/Power Meter( PM) 硬度計/Sclerometer( S) 秒表/Stop Watch( SW) 刀口尺/Knife Gauge( KG) 放大鏡/Amlifer( A) 水平儀/Level Balance ( LB) 內徑千分尺/Internal Micrometer( IM) 卷尺/Roll Meter( RM) 耐壓測試儀器/Pressure Meter( PM) 千分尺/Micrometer( M) 內側千分尺/Insiele Micrometer( IM) 深度千分尺/Depth Micrometer( DM)回复 引用 TOP发短消息 加为好友hai 当前离线 UID 9帖子 439 精华 6积分 596 威望 81 金钱 184 阅读权限 50在线时间 40 小时 注册时间2009-8-6 最后登录2010-10-11 3# hai 发表于 2009-8-614:29 | 只看该作者踩窝窝送礼物 问候Ta 塑料模专业英语 1、ejector unit 顶出单元,包括一切有顶出功能的零件:ejector pin, ejector plate,ejector sleeve,ejector rod, ejector leader busher 顶出导销(顶出板导杆)的衬套,也叫ejector guide bush ejector stopper,用于顶出制动的,或限位的 ejector pin retaining plate:顶针固定板。
Hybrid
计算机科技与发展Hybrid Genetic Algorithm Based Image Enhancement TechnologyMu Dongzhou Department of the Information Engineering XuZhou College of Industrial Technol ogyXuZhou, China mudzh@Xu Chao and Ge Hongmei Department of the Information Engineering XuZhou College of Indus trial TechnologyXuZhou, China xuch@ , gehm@Abstract-in image enhancement, Tubbs proposed a normalized incomplete Beta function to represe nt several kinds of commonly used non-linear transform functions to do the research on image enh ancement. But how to define the coefficients of the Beta function is still a problem. We proposed a Hybrid Genetic Algorithm which combines the Differential Evolution to the Genetic Algorithm in the image enhancement process and utilize the quickly searching ability of the algorithm to carry out the adaptive mutation and searches. Finally we use the Simulation experiment to prove the effe ctiveness of the method.Keywords- Image enhancement; Hybrid Genetic Algorithm; adaptive enhancementI. INTRODUCTIONIn the image formation, transfer or conversion process, due to other objective factors such as syste m noise, inadequate or excessive exposure, relative motion and so the impact will get the image of ten a difference between the original image (referred to as degraded or degraded) Degraded image is usually blurred or after the extraction of information through the machine to reduce or even wro ng, it must take some measures for its improvement.Image enhancement technology is proposed in this sense, and the purpose is to improve the image quality. Fuzzy Image Enhancement situation according to the image using a variety of special tech nical highlights some of the information in the image, reduce or eliminate the irrelevant informatio n, to emphasize the image of the whole or the purpose of local features. Image enhancement meth od is still no unified theory, image enhancement techniques can be divided into three categories: p oint operations, and spatial frequency enhancement methods Enhancement Act. This paper present s an automatic adjustment according to the image characteristics of adaptive image enhancement method that called hybrid genetic algorithm. It combines the differential evolution algorithm of ad aptive search capabilities, automatically determines the transformation function of the parameter v alues in order to achieve adaptive image enhancement.Digital Image ProcessingInterest in digital image processing methods stems from two principal applica- tion areas: impro vement of pictorial information for human interpretation; and processing of image data for stor age, transmission, and representation for au- tonomous machine perception.An image may be defined as a two-dimensional function, f(x, y), where x and y are spatial (plane ) coordinates, and the amplitude of f at any pair of coordinates (x, y) is called the intensity or gra y level of the image at that point. When x, y, and the amplitude values of f are all finite, discrete quantities, we call the image a digital image. The field of digital image processing refers to proces sing digital images by means of a digital computer. Note that a digital image is composed of a fini te number of elements, each of which has a particular location and value. These elements are ref erred to as picture elements, image elements, pels, and pixels. Pixel is the term most widely used t o denote the elements of a digital image.Vision is the most advanced of our senses, so it is not surprising that images play the single most important role in human perception. However, unlike humans, who are limited to the visual ban d of the electromagnetic (EM) spec- trum, imaging machines cover almost the entire EM spect rum, ranging from gamma to radio waves. They can operate on images generated by sources tha t humans are not accustomed to associating with images. These include ultra- sound, electron microscopy, and computer-generated images. Thus, digital image processing encompasses a wi de and varied field of applications.There is no general agreement among authors regarding where image processing stops and othe r related areas, such as image analysis and computer vi- sion, start. Sometimes a distinction is made by defining image processing as a discipline in which both the input and output of a proces s are images. We believe this to be a limiting and somewhat artificial boundary. For example, und er this definition, even the trivial task of computing the average intensity of an image (which yi elds a single number) would not be considered an image processing operation. On the other han d, there are fields such as computer vision whose ultimate goal is to use computers to emulate h uman vision, including learning and being able to make inferences and take actions based on vis ual inputs. This area itself is a branch of artificial intelligence (AI) whose objective is to emulat e human intelligence. The field of AI is in its earliest stages of infancy in terms of development, with progress having been much slower than originally anticipated. The area of image analysis (a lso called image understanding) is in be- tween image processing and computer vision.There are no clearcut boundaries in the continuum from image processing at one end to compute r vision at the other. However, one useful paradigm is to consider three types of computerized processes in this continuum: low-, mid-, and highlevel processes. Low-level processes involve p rimitive opera- tions such as image preprocessing to reduce noise, contrast enhancement, and i mage sharpening. A low-level process is characterized by the fact that both its inputs and outpu ts are images. Mid-level processing on images involves tasks such as segmentation (partitionin g an image into regions or objects), description of those objects to reduce them to a form suit able for computer processing, and classification (recognition) of individual objects. A midlevel process is characterized by the fact that its inputs generally are images, but its outputs are attribut es extracted from those images (e.g., edges, contours, and the identity of individual objects). Fi nally, higherlevel processing involves “making sense” of an ensemble of recognized objects, as in image analysis, and, at the far end of the continuum, performing the cognitive functions n ormally associated with vision.Based on the preceding comments, we see that a logical place of overlap between image processi ng and image analysis is the area of recognition of individual regions or objects in an image. T hus, what we call in this book digital image processing encompasses processes whose inputs and outputs are images and, in addition, encompasses processes that extract attributes from images, up to and including the recognition of individual objects. As a simple illustration to clarify these concepts, consider the area of automated analysis of text. The processes of acquiring an imag e of the area containing the text, preprocessing that image, extracting (segmenting) the individu al characters, describing the characters in a form suitable for computer processing, and recogn izing those individual characters are in the scope of what we call digital image processing in this book. Making sense of the content of the page may be viewed as being in the domain of image a nalysis and even computer vision, depending on the level of complexity implied by the statemen t “making sense.” As will become evident shortly, digital image processing, as we have define d it, is used successfully in a broad range of areas of exceptional social and economic value.The areas of application of digital image processing are so varied that some form of organization is desirable in attempting to capture the breadth of this field. One of the simplest ways to devel op a basic understanding of the extent of image processing applications is to categorize images according to their source (e.g., visual, X-ray, and so on). The principal energy source for images i n use today is the electromagnetic energy spectrum. Other important sources of energy include acoustic, ultrasonic, and electronic (in the form of electron beams used in electron microscopy). Synthetic images, used for modeling and visualization, are generated by computer. In this sectio n we discuss briefly how images are generated in these various categories and the areas in which t hey are applied.Images based on radiation from the EM spectrum are the most familiar, es-pecially images in the X-ray and visual bands of the spectrum. Electromagnet- ic waves can be conceptualized as propa gating sinusoidal waves of varying wavelengths, or they can be thought of as a stream of massl ess particles, each traveling in a wavelike pattern and moving at the speed of light. Each massless particle contains a certain amount (or bundle) of energy. Each bundle of energy is called a pho ton. If spectral bands are grouped according to energy per photon, we obtain the spectrum sh own in fig. below, ranging from gamma rays (highest energy) at one end to radio waves (lowest energy) at the other. The bands are shown shaded to convey the fact that bands of the EM spectr um are not distinct but rather transition smoothly from one to the other.Image acquisition is the first process. Note that acquisition could be as simple as being given an i mage that is already in digital form. Generally, the image acquisition stage involves preprocessin g, such as scaling.Image enhancement is among the simplest and most appealing areas of digital image processing. Basically, the idea behind enhancement techniques is to bring out detail that is obscured, or sim ply to highlight certain features of interest in an image. A familiar example of enhancement is w hen we increase the contrast of an image because “it looks better.” It is important to keep in m ind that enhancement is a very subjective area of image processing. Image restoration is an area t hat also deals with improving the appearance of an image. However, unlike enhancement, whichis subjective, image restoration is objective, in the sense that restoration techniques tend to be based on mathematical or probabilistic models of image degradation. Enhancement, on the oth er hand, is based on human subjective preferences regarding what constitutes a “good” enha ncement result.Color image processing is an area that has been gaining in importance because of the significant increase in the use of digital images over the Internet. It covers a number of fundamental conce pts in color models and basic color processing in a digital domain. Color is used also in later cha pters as the basis for extracting features of interest in an image.Wavelets are the foundation for representing images in various degrees of resolution. In partic ular, this material is used in this book for image data compression and for pyramidal representati on, in which images are subdivided successively into smaller regions.Compression, as the name implies, deals with techniques for reducing the storage required to save an image, or the bandwidth required to transmi it.Although storage technology has improved sign ificantly over the past decade, the same cannot be said for transmission capacity. This is true parti cularly in uses of the Internet, which are characterized by significant pictorial content. Image c ompression is familiar (perhaps inadvertently) to most users of computers in the form of image file extensions, such as the jpg file extension used in the JPEG (Joint Photographic Experts Gro up) image compression standard.Morphological processing deals with tools for extracting image components that are useful in the representation and description of shape. The material in this chapter begins a transition from p rocesses that output images to processes that output image attributes.Segmentation procedures partition an image into its constituent parts or objects. In general, auto nomous segmentation is one of the most difficult tasks in digital image processing. A rugged seg mentation procedure brings the process a long way toward successful solution of imaging proble ms that require objects to be identified individually. On the other hand, weak or erratic segment ation algorithms almost always guarantee eventual failure. In general, the more accurate the seg mentation, the more likely recognition is to succeed.Representation and description almost always follow the output of a segmentation stage, which usually is raw pixel data, constituting either the bound- ary of a region (i.e., the set of pixels sepa rating one image region from another) or all the points in the region itself. In either case, converti ng the data to a form suitable for computer processing is necessary. The first decision that must be made is whether the data should be represented as a boundary or as a complete region. Bo undary representation is appropriate when the focus is on external shape characteristics, suc h as corners and inflections. Regional representation is appropriate when the focus is on in ternal properties, such as texture or skeletal shape. In some applications, these representations complement each other. Choosing a representation is only part of the solution for trans- forming raw data into a form suitable for subsequent computer processing. A method must also be specif ied for describing the data so that features of interest are highlighted. Description, also called feature selection, deals with extracting attributes that result in some quantitative information of in terest or are basic for differentiating one class of objects from another.Recognition is the process that assigns a label (e.g., “vehicle”) to an object based on its descr iptors. As detailed before, we conclude our coverage of digital image processing with the develo pment of methods for recognition of individual objects.So far we have said nothing about the need for prior knowledge or about the interaction betwe en the knowledge base and the processing modules in Fig2 above. Knowledge about a problem domain is coded into an image processing system in the form of a knowledge database. This kn owledge may be as sim- ple as detailing regions of an image where the information of interest is known to be located, thus limiting the search that has to be conducted in seeking that information. The knowledge base also can be quite complex, such as an interrelated list of all major possible defects in a materials inspection problem or an image database containing high-resolution sate llite images of a region in con- nection with change-detection applications. In addition to guidi ng the operation of each processing module, the knowledge base also controls the interaction be tween modules. This distinction is made in Fig2 above by the use of double-headed arrows betw een the processing modules and the knowledge base, as op- posed to single-headed arrows linkin g the processing modules.II. IMAGE ENHANCEMENT TECHNOLOGYImage enhancement refers to some features of the image, such as contour, contrast, emphasis or hi ghlight edges, etc., in order to facilitate detection or further analysis and processing. Enhancement s will not increase the information in the image data, but will choose the appropriate features of th e expansion of dynamic range, making these features more easily detected or identified, for the det ection and treatment follow-up analysis and lay a good foundation.Image enhancement method consists of point operations, spatial filtering, and frequency domain fi ltering categories. Point operations, including contrast stretching, histogram modeling, and limitin g noise and image subtraction techniques. Spatial filter including low-pass filtering, median filteri ng, high pass filter (image sharpening). Frequency filter including homomorphism filtering, multi-scale multi-resolution image enhancement applied.III. DIFFERENTIAL EVOLUTION ALGORITHMDifferential Evolution (DE) was first proposed by Price and Storn, and with other evolutionary alg orithms are compared, DE algorithm has a strong spatial search capability, and easy to implement, easy to understand. DE algorithm is a novel search algorithm, it is first in the search space random ly generates the initial population and then calculate the difference between any two members of t he vector, and the difference is added to the third member of the vector, by which Method to form a new individual. If you find that the fitness of new individual members better than the original, th en replace the original with the formation of individual self.The operation of DE is the same as genetic algorithm, and it conclude mutation, crossover and sele ction, but the methods are different. We suppose that the group size is P, the vector dimension is D,and we can express the object vector as (1):xi=[xi1,xi2,…,xiD] (i =1,…,P) (1)And the mutation vector can be expressed as (2):Vi=Xrl+F*(Xr2-X1) (2)Xrl,Xr2,Xr3are three randomly selected individuals from group, and r1≠r2≠r3≠i.F is a range of [0, 2] between the actual type constant factor difference vector is used to control the influence, co mmonly referred to as scaling factor. Clearly the difference between the vector and the smaller the disturbance also smaller, which means that if groups close to the optimum value, the disturbance will be automatically reduced.DE algorithm selection operation is a "greedy " selection mode, if and only if the new vector ui th e fitness of the individual than the target vector is better when the individual xi, ui will be retained to the next group. Otherwise, the target vector xi individuals remain in the original group, once ag ain as the next generation of the parent vector.IV. EXPERIMENT AND ANAL YSISIn the simulation, we used two different types of gray-scale images degraded; the program perfor med 50 times, population sizes of 30, evolved 600 times. The results show that the proposed meth od can very effectively enhance the different types of degraded image.Figure 2, the size of the original image a 320 × 320, it's the contrast to low, and some details of th e more obscure, in particular, scarves and other details of the texture is not obvious, visual effects, poor, using the method proposed in this section, to overcome the above some of the issues and get satisfactory image results, as shown in Figure 5 (b) shows, the visual effects have been well impro ved. From the histogram view, the scope of the distribution of image intensity is more uniform, an d the distribution of light and dark gray area is more reasonable. Hybrid genetic algorithm to auto matically identify the nonlineartransformation of the function curve, and the values obtained before 9.837,5.7912, from the curve can be drawn, it is consistent with Figure 3, c-class, that stretch across the middle region compress ion transform the region, which were consistent with the histogram, the overall original image low contrast, compression at both ends of the middle region stretching region is consistent with huma n visual sense, enhanced the effect of significantly improved.Figure 3, the size of the original image a 320 × 256, the overall intensity is low, the use of the me thod proposed in this section are the images b, we can see the ground, chairs and clothes and other details of the resolution and contrast than the original image has Improved significantly, the origin al image gray distribution concentrated in the lower region, and the enhanced image of the gray un iform, gray before and after transformation and nonlinear transformation of basic graph 3 (a) the s ame class, namely, the image Dim region stretching, and the values were 5.9409,9.5704, nonlinear transformation of images degraded type inference is correct, the enhanced visual effect and good r obustness enhancement.Difficult to assess the quality of image enhancement, image is still no common evaluation criteria, common peak signal to noise ratio (PSNR) evaluation in terms of line, but the peak signal to nois e ratio does not reflect the human visual system error. Therefore, we use marginal protection index and contrast increase index to evaluate the experimental results.In figure 4, we compared with the Wavelet Transform based algorithm and get theevaluate number in TABLE I.Figure 4 (a, c) show the original image and the differential evolution algorithm for enhanced result s can be seen from the enhanced contrast markedly improved, clearer image details, edge feature more prominent. b, c shows the wavelet-based hybrid genetic algorithm-based Comparison of Ima ge Enhancement: wavelet-based enhancement method to enhance image detail out some of the im age visual effect is an improvement over the original image, but the enhancement is not obvious; a nd Hybrid genetic algorithm based on adaptive transform image enhancement effect is very good, image details, texture, clarity is enhanced compared with the results based on wavelet transform h as greatly improved the image of the post-analytical processing helpful. Experimental enhancemen t experiment using wavelet transform "sym4" wavelet, enhanced differential evolution algorithm e xperiment, the parameters and the values were 5.9409,9.5704. For a 256 × 256 size image transfo rm based on adaptive hybrid genetic algorithm in Matlab 7.0 image enhancement software, the co mputing time is about 2 seconds, operation is very fast. From TABLE I, objective evaluation criter ia can be seen, both the edge of the protection index, or to enhance the contrast index, based on ad aptive hybrid genetic algorithm compared to traditional methods based on wavelet transform has a larger increase, which is from This section describes the objective advantages of the method. Fro m above analysis, we can see that this method.From above analysis, we can see that this method can be useful and effective.V. CONCLUSIONIn this paper, to maintain the integrity of the perspective image information, the use of Hybrid gen etic algorithm for image enhancement, can be seen from the experimental results, based on the Hy brid genetic algorithm for image enhancement method has obvious effect. Compared with other ev olutionary algorithms, hybrid genetic algorithm outstanding performance of the algorithm, it is si mple, robust and rapid convergence is almost optimal solution can be found in each run, while thehybrid genetic algorithm is only a few parameters need to be set and the same set of parameters ca n be used in many different problems. Using the Hybrid genetic algorithm quick search capability for a given test image adaptive mutation, search, to finalize the transformation function from the b est parameter values. And the exhaustive method compared to a significant reduction in the time to ask and solve the computing complexity. Therefore, the proposed image enhancement method has some practical value.基于混合遗传算法的图像增强技术Mu Dongzhou 徐州工业职业技术学院信息工程系XuZhou, China mudzh@Xu Chao and Ge Hongmei 徐州工业职业技术学院信息工程系XuZhou, China xuch@ , gehm@摘要—在图像增强之中,塔布斯提出了归一化不完全β函数表示常用的几种使用的非线性变换函数对图像进行研究增强。
Immune genetic algorithm for weapon-target assignment problem
Chinese Academy of Sciences, Beijing 100080,China E-mail:gao_shang@
k =1
j =1
i=1
K
∑ s.t. nk = T
(2)
k =1
T
∑ xij = 1 , i = 1,2,",W
(3)
j =1
xij ∈ {0,1}, i = 1,2,",W , j = 1,2,",T (4)
where xij is a Boolean value indicating whether the
Abstract
An Immune Genetic Algorithm (IGA) is used to solve weapon-target assignment problem (WTA). The used immune system serves as a local search mechanism for genetic algorithm. Besides, in our implementation, a new crossover operator is proposed to preserve good information contained in the chromosome. A comparison of the proposed algorithm with several existing search approaches shows that the IGA outperforms its competitors on all tested WTA problems.
基于混合整数规划的路径规划算法研究
基于混合整数规划的路径规划算法研究混合整数规划(Mixed Integer Programming,简称MIP)作为一种强大的数学优化工具,广泛应用于各个领域。
路径规划作为其中的一个重要应用之一,旨在寻找最佳路径以满足特定的约束条件和目标函数。
本文将围绕基于混合整数规划的路径规划算法展开研究,旨在探究该算法的原理和应用。
1. 引言路径规划是计算机科学和运筹学领域的一个重要研究问题,它在实际应用中有着广泛的需求。
在许多场景中,我们需要在给定的网络结构中找到最佳路径,以达到一定的目标。
而基于混合整数规划的路径规划算法,通过将路径规划问题转化为一种约束优化问题求解,可以有效地得到最优解。
2. 混合整数规划简介混合整数规划是一类特殊的线性规划问题,在该问题中,变量可以是实数或整数。
其数学模型可以表示为:```max/min C*Xs.t. A*X ≤ bX ∈ Z^m x R^n```其中,C是目标函数的系数矩阵,A是约束条件的系数矩阵,b是约束条件右侧的常数向量,X是决策变量向量。
混合整数规划在数学上是一个NP-hard问题,求解该问题需要借助于专门的算法。
3. 基于混合整数规划的路径规划算法基于混合整数规划的路径规划算法的核心思想是将路径规划问题转化为一个混合整数规划问题,并通过求解该问题获得最佳路径。
该算法主要包括以下几个步骤:步骤1:建立混合整数规划模型。
根据路径规划问题的具体特点,将起点、终点、障碍物等信息转化为约束条件和目标函数,构建混合整数规划模型。
步骤2:求解混合整数规划模型。
利用现有的优化求解算法,对混合整数规划模型进行求解,得到最优解。
步骤3:提取最优路径。
根据混合整数规划模型的求解结果,提取最优路径并进行输出,作为最终的路径规划结果。
4. 算例分析为了验证基于混合整数规划的路径规划算法的有效性,我们选取了一个实际问题进行了算例分析。
在一个城市中,有多个货物需要从起点运送到目的地,货车的路径规划问题成为一个关键的调度问题。
A Genetic Algorithm for the Multi-Pickup and Delivery Problem with time windows
The PDPTW is a variant of VRPTW where in addition to the existence of time constraints, this problem implies a set of customers and a set of suppliers geographically located. Every routing must also satisfy the precedence constraints to ensure that a customer should not be visited before his supplier. [Psaraftis, H.N., 1983] A dynamic approach for resolve the 1-PDP without and with time windows was developed by Psaraftis, H.N considering objective function as a minimization weighting of the total travel time and the non-customer satisfaction. [Psaraftis, H.N., 1980] Jih, W and al have developed an approach based on the hybrid genetic algorithms to solve the 1PDPTW, aiming to minimize combination of the total cost and total waiting time. [Jih, W and al., 1999] Another genetic algorithm was developed by Velasco, N and al to solve the 1-PDP bi-objective in which the total travel time must be minimized while satisfy in prioritise the most urgent requests. In this literature, the method proposed to resolve this problem is based on a No dominated Sorting Algorithm (NSGA-II). [Velasco, N and al., 2006] Kammarti, R and al deal the 1-PDPTW, minimizing the compromise between the total travel distance, total waiting time and total tardiness time, using an evolutionary algorithm with Special genetic operators, tabu search to provide a set of viable solutions. [Kammarti, R and al., 2004] [Kammarti, R and al 2005a] This work have been extended, in proposing a new approach based on the use of lower bounds and Pareto dominance method, to minimize the compromise between the total travel distance and total tardiness time. [Kammarti, R and al 2006] [Kammarti, R and al 2007] About the m-PDPTW, Sol, M and al have proposed a branch and price algorithm to solve the mPDPTW, minimizing the vehicles number required to satisfy all travel demands and the total travel distance. [Sol, M and al., 1994] Quan, L and al have presented a construction heuristic based on the integration principle with the objective function, minimizing the total cost, including the vehicles fixed costs and travel expenses that are proportional to the travel distance. [Quan, L and al., 2003] A new metaheuristic based on a tabu algorithm, was developed by Li, H and al to solve the mPDPTW. [Li, H and al., 2001] Li, H and al have developed a “Squeaky wheel” method to solve the m-PDPTW with a local search. [Li, H and al., 2002] A genetic algorithm was developed by Harbaoui Dridi, I and al dealing the m-PDPTW to minimize the total travel distance and the total transport cost. [Harbaoui Dridi, I and al., 2008]
基于自适应MPGA的湿度敏感类物品立体仓库货位分配方法
基于自适应MPGA的湿度敏感类物品立体仓库货位分配方法湿度敏感类物品的存储和保管是立体仓库管理中的一个重要问题。
湿度对这类物品的质量和保质期有着直接的影响,因此,科学合理地分配货位,使湿度敏感类物品得到最佳的存储条件,是一个具有挑战性的任务。
本文提出了一种基于自适应多目标遗传算法(MPGA)的湿度敏感类物品立体仓库货位分配方法,以实现对湿度敏感类物品的有效管理。
1. 引言立体仓库作为一种高效的货物存储和管理方式,广泛应用于物流和供应链领域。
湿度敏感类物品的存储需要特殊的环境条件,因此,在立体仓库中的货位分配是非常重要的。
传统的货位分配方法往往无法满足湿度敏感类物品的要求,因此需要引入自适应MPGA方法来优化货位的分配。
2. 自适应多目标遗传算法(MPGA)自适应多目标遗传算法是一种基于进化计算的优化算法,具有对多个目标进行优化的能力,并且可以根据问题的特点进行自适应调整。
自适应MPGA将货位分配问题转化为一个多目标优化问题,并通过不断的进化来找到最优的解。
3. 湿度敏感类物品的特点湿度敏感类物品对湿度的要求较高,过高或过低的湿度都会对物品的质量产生负面影响。
因此,在货位分配时需要考虑湿度的影响因素,并将湿度作为一个优化目标。
4. 基于自适应MPGA的湿度敏感类物品货位分配方法在基于自适应MPGA的湿度敏感类物品货位分配方法中,首先需要定义适应度函数。
适应度函数将考虑物品的湿度要求、存储效率以及货位的可用性等因素。
通过不断进化操作,找到最优的货位分配方案。
5. 实验结果与分析通过对实际的湿度敏感类物品仓库进行模拟实验,验证了基于自适应MPGA的货位分配方法的有效性。
实验结果显示,该方法可以有效地降低湿度敏感类物品的损失,并提高货位的利用率。
6. 结论本文提出了一种基于自适应MPGA的湿度敏感类物品立体仓库货位分配方法。
该方法通过考虑湿度要求、存储效率和货位可用性等因素,优化湿度敏感类物品的货位分配方案。
基因组装方法
基因组装方法Genome assembly is a fundamental process in bioinformatics that involves piecing together the DNA sequences of an organism. This can be a complex and daunting task due to the sheer amount of data involved and the presence of repetitive sequences in the genome. 基因组装是生物信息学中的一个基本过程,涉及将一个有机体的DNA序列拼接在一起。
这可能是一个复杂而令人生畏的任务,因为涉及的数据量庞大,并且基因组中存在重复序列。
One common method used for genome assembly is known as de novo assembly. This approach involves reconstructing the genome from scratch without the aid of a reference genome. While de novo assembly can be more challenging compared to reference-guided assembly, it is particularly useful for non-model organisms or species with complex genomes. 一个常用的基因组装方法是所谓的de novo组装。
这种方法涉及在没有参考基因组的帮助下从头开始重建基因组。
虽然与参考引导组装相比,de novo组装可能更具挑战性,但对于非模式生物或基因组复杂的物种特别有用。
Another important aspect of genome assembly is the choice of sequencing technology. Different sequencing platforms such as Illumina, PacBio, and Oxford Nanopore have different advantagesand limitations when it comes to generating DNA sequences. The selection of the appropriate sequencing technology can greatly impact the quality and accuracy of the assembled genome. 基因组装的另一个重要方面是选择测序技术。
基于遗传-贪心混合搜索的人造板下料算法
128
林业工程学报
第6卷
in this kind of layout calculation. Therefore, the utilization rate of raw materials can be further improved. Keywords:genetic algorithm; greedy algorithm; cutting of wood⁃based panel; guillotine
林业工程学报, 2021,6(4) :127-133 Journal of Forestry Engineering
DOI:10.13360 / j.issn.2096-1359.202010037
基于遗传⁃贪心混合搜索的人造板下料算法
刘诚,孙远升,花军∗,姚嘉明
( 东北林业大学机电工程学院,哈尔滨 150040)
显示:在非“ 一刀切” 排样算例计算中,相比文献算法平均排样效果提高了 2%,而计算时间缩减 90%,相比文献
同类型算法在获得相同求解效果时,迭代次数缩减 50%;在有“ 一刀切” 约束排样算例计算中,相比文献算法平
均排样效果提高了 2.7%,而平均余料数量减少了 50%。 试验证明:在进行非“ 一刀切” 排样计算时,具有较高的
智能算法也大量应用在矩形排样问题求解中, 其中 遗 传 算 法 应 用 最 为 广 泛, 如 Aurelio、 赵 新 芳 等[10-11] 均在无“ 一刀切” 约束的矩形排样中使用遗 传算法,获得较高排样利用率,但这类算法求解具 有随机性,收敛速度及求解效果较为依赖遗传策略 的选择。
针对上述算法存在问题,笔者采取对排样择优 后进行“一刀切” 检测的排样策略,降低余料碎化 程度,通过改进遗传算法进行全局搜索,之后利用 贪心算法对其进行二次优化,使解的质量接近全局 最优,提高排样利用率,为人造板优选下料提供一 种新的解决方案。
求解混合流水车间调度的改进贪婪遗传算法
第 41 卷 第 5 期
系统工程与电子技术
Vol.41 No.5
2019 年 5 月
SystemsEngineeringandElectronics
May2019
文 章 编 号 :1001506X(2019)05107908
网 址 :www.sysele.com
求解混合流水车间调度的改进贪婪遗传算法
HFSP 按并行机类型 可 分 为 3 类[1]:① 相 同 并 行 机 HFSP, 即每一阶段上同一工件在任一台并行设备上的加工时间 相 同 ;② 均 匀 并 行 机 HFSP,即 每 一 阶 段 上 同 一 工 件 在 任 一台并行设备上的加工时间与该台设备的加工速度成反 比 ;③不相关并行机 HFSP,即每 一 阶 段 上 同 一 工 件 在 任 两 台并行设备上的加工 时 间 互 不 相 关,而 取 决 于 工 件 与 设 备 的匹配程度。
犓犲狔狑狅狉犱狊:hybridflowshopschedulingproblem;greedygeneticalgorithm;forwardandreversedecoding; makespan
0 引 言
混合流 水 车 间 调 度 问 题 (hybridflowshopscheduling problem,HFSP)是在 传 统 流 水 车 间 调 度 问 题 基 础 上,存 在 多个阶段的加工设备配置了多台的情况。因此该调度问题 就由工件排序和设备 分 配 两 个 工 作 组 成,是 传 统 流 水 车 间 调度与并行机调 度 的 综 合。 相 比 流 水 车 间 调 度 问 题,其 复 杂度更高,求解空 间 更 大。 同 时 该 问 题 具 有 很 强 的 工 业 背 景,广 泛 存 在 于 机 械 加 工、冶 炼、食 品 加 工、化 工 等 行 业,因 此,对此问题进行 研 究 具 有 重 要 的 学 术 价 值 和 意 义。 传 统
基于遗传模糊系统的兵棋推演关键点推理方法
犃犫狊狋狉犪犮狋:Inrecentyears,moreand moreattentionhasbeenpaidtoautomaticcombatdeductionbasedon artificialintelligencetechnology.However,duetothedifficultyincollectingcombatdeductiondataeffectively, manyartificialintelligencetechnologiesthatrelyondatalearningareineffective.Itisafeasiblealternativeto studytheartificialintelligencetechnologyincombatdeductionbasedonexpertknowledgeandcombatdeduction data.Therefore,basedonthewargame,ageneticfuzzysystem (GFS)frameworkforkeypointsreasoningis designed,whicheffectivelyintegratestheknowledgemodelingofwargameexpertsandthelearningofwargame replaydata,thusimprovingthereasoningqualityofkeypoints.Andthen,takingsecuringpointsasanexample,a GFSforsecuringpointreasoningisconstructed.Onthebasisofinitializingtherulebaseofthefuzzysystem,bymeans ofthereasonabledesignofparametercoding,fitnessfunctionandgeneticoperatorinthegeneticalgorithm,thege netictuningalgorithmforsecuringpointreasoningfuzzysystemisimplemented.Finally,thefeasibilityand practicabilityoftheproposedmodelaredemonstratedthroughtheexperimentalsimulation.
基于混合蚂蚁算法的二维装箱问题求解
收稿日期:2003-11-14;修订日期:2004-01-09 作者简介:赵中凯(1978-),男,甘肃白银人,助教,硕士,主要研究方向:智能优化、计算机算法设计; 梅国建(1955-),男,河南许昌人,副教授,主要研究方向:管理科学、系统工程; 沈洪(1971-),男,湖南澧县人,讲师,主要研究方向:管理科学、系统工程; 赵战彪(1972-),男,河北博野县人,讲师,硕士,主要研究方向:系统优化.文章编号:1001-9081(2004)06Z -0297-02基于混合蚂蚁算法的二维装箱问题求解赵中凯,梅国建,沈 洪,赵战彪(装甲兵工程学院装备管理室,北京100072)摘 要:二维装箱问题是一个NP -Hard 组合优化问题。
根据蚂蚁优化算法和二维装箱问题的特点,本文提出了改进的BL 算法与蚂蚁算法相结合的混合算法来解决二维装箱问题,实验结果表明,该算法是行之有效的,并具有一定的通用性。
关键词:二维装箱;蚂蚁算法;BL 算法;优化中图分类号:TP301 文献标识码:A1 引言在许多工业领域,人们需要把矩形物体装入诸如集装箱之类的容器中,例如在码头,大小重量不一的集装箱需要装入货船中;在木材或钢铁厂中,根据实际需要从一个大的原材料中切割成不同形状的几块部分,进行分别包装处理;在超市中尤其是仓储超市中,各种商品需要分类放置在不同的货架上。
在所有类似的应用中,如何最有效的利用容器空间,按照需要的次序放置物体,是此类问题研究的目的。
20世纪80年代以来,许多学者尝试应用各种方法解决该类问题,其中遗传算法(G A )、模拟退火算法(S A )等现代优化算法在该问题中得到了广泛的应用。
但是由于实际问题约束条件的复杂性,启发式算法的研究比较引人关注[1]。
1980图1 BL 改进算法流程图年,Jacos 提出Bottom 2Left (BL )算法,由于该算法利用矩形的长度和宽度进行物体放置,要求被放置物体尽可能快地到达容器底部的同时尽可能快地到达容器的左侧,不用确定放置点的坐标,解决了物体装入过程中图形结果的输出问题和物体之间的干涉(重叠)问题,因而得到了广泛的研究。
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
ISSN1746-7659, England, UKJournal of Information and Computing ScienceVol. 1, No. 1, 2006, pp. 61-64A Hybrid Genetic Algorithm for Bin Packing ProblemBased on Item SequencingXianghe Jing1, 2 + Xianzhong Zhou 3 Yangyong Xu 21 Dept. of Automation, Nanjing University of Science & Technology, Nanjing 210094, China2 Dept. of Information & Control, Air Defence Command College, Zhengzhou 450052, China3 School of Management & Engineering, Nanjing University, Nanjing 210093, China( Received Sep. 14, 2005, accepted Dec. 04, 2005 )Abstract. The bin packing problem based on item sequencing is defined as follows: given a set of sequenced items of different sizes, how should one pack them all into bins of different capacity, in order to make the utility ratio of bin capacity as higher as possible? In this paper, next fit algorithm, a heuristic method for bin packing problem, is introduced into simple genetic algorithm, and a hybrid genetic algorithm is proposed for solving bin-packing problem based on item sequencing. In the hybrid genetic algorithm, the idea of simple genetic algorithm is used to search the solution of the bins sequence, and the idea of next fit algorithm is used to pack the sequenced items into the bins sequence obtained. Finally, the effectiveness of the hybrid genetic algorithm is convinced through computational results of an example.Keywords: genetic algorithm, next fit algorithm, bin packing, item sequencing.1.IntroductionThe bin-packing problem is an NP-completeness problem [1]. The one-dimensional bin-packing problem based on item sequencing is defined as follows: given a set of M sequenced items {I1, I2, …, I M} of different sizes {S1, S2, …, S M}, and different kinds of bins {B1, B2, …, B N} of different capacities {C1, C2, …, C N}. The problem is how to pack all the items into the bins so that the total size of items in any bin must not exceed its capacity and the position of the sequenced items cannot be changed and the utility ratio of bin capacity is as higher as possible.Since genetic algorithm (GA) was first introduced some 30 years ago [2], it has been widely recognized as an effective search paradigm in artificial intelligence, image processing, job scheduling, pattern recognition and many other areas, particularly as a method for gaining relatively good solutions to NP-hard optimization problems [3]. In references [4 ~ 7], some methods for solving bin-packing problem have been studied. In this paper, a hybrid genetic algorithm (HGA) is proposed for solving one-dimensional bin-packing problem based on item sequencing.2.Design of hybrid genetic algorithmIn bin packing problem based on item sequencing, there are two key factors that influence the solution of the problem. The first factor is how to select and sequence the different capacity bins. The other factor is how to pack the sequenced item into the bin sequence. Therefore, in our HGA, the idea of GA is used to search the solution of the bin sequence, and the idea of next fit algorithm (NF) [8] is used to pack the sequenced items into the bin sequence obtained.2.1.NF procedureNF is a heuristic method for bin packing problem. On the assumption that bins have been selected and sequenced and stored in a list L, the k-th bin number in the list L is denoted as L k(∈{1, 2, …, N *}, N*≤M). Then we can pack the given sequenced items into the bin sequence. In NF procedure, the items are packed in the following way.+ Corresponding author. Tel.: +86-25-84303047-8;E-mail address: luckycrane@Published by World Academic Press, World Academic UnionX. Jing, et al: A Hybrid Genetic Algorithm for Bin Packing Problem Based on Item Sequencing62 Step 1: Set m ←1, n ←1. The variable i and j represent the item number and the bin number, respectively. Step 2: If m ≤ M, turn to step 3. Otherwise, turn to step 5.Step 3: Let SI k be the set of items in bin L k . If S m +∑n*∈SIn S n* ≤C n , pack the item I m into the bin L n and set m ←m+1 and return to step 2. Otherwise, close the bin L n and turn to step 4.Step 4: If n <N *, set n ←n+1 and return to step 2. Otherwise, select a new bin and add it into the end of the bin sequence, set N *←N *+1, n ←n+1, and return to step 2.Step 5: Terminate the procedure.2.2. Individual descriptionIn our HGA, the kind number of the bins corresponds to a gene. The chromosome is expressed by a set of sequence numbers. For example, a chromosome444344421K K mk c 2152)(= It means that the kind of the first and the m-th bin in the list L is B 2, and that the kind of the second bin in the list L is B 5, and that the kind of the third bin in the list L is B 1. From the chromosome, we can obtain the kinds of the bins used in the packing problem. The initial solutions are a set of N (N ≤M ) numbers, generated from 1 to N randomly.* *2.3. Fitness of algorithmIn the packing problem, our aim is to make the utility ratio of bin capacity as higher as possible. Let Tn be the total number of bins that are used in a solution after packing by NF procedure, and Ts (I ) be the total size of the items, and Tc (B ) be the total capacity of the bins used in a solution. We define the objective function as follows.∑∑====Tn n n M m m C S B Tc I Ts X f 11/)(/)()(min Let E (X ) be the fitness function of HGA. Because the value of f (X ) is positive, we look f (X ) as the fitness function, that is to say,E (X ) = f (X )In the algorithm, the larger the value of fitness function is, the better the fitness is, and the better the solution is.2.4. Selection operatorThe selection operator should be designed in such a way that better individuals can be copied to offspring. In simple genetic algorithm, the selection operation depends on the ratio of individual fitness to average fitness, which cannot show the ascendant of better individuals when the fitness of every individual is almost the same size. And this phenomenon may reflect premature convergence problem of genetic algorithm. So we present our following design of selection operator.Step 1: Sort the individuals in order of non-increasing value of fitness.Step 2: Make two copies from the first 1/3 part of the sequenced individuals.Step 3: Make one copy from the second 1/3 part of the sequenced individuals.Step 4: Make above copies be the new individuals (offspring).The design of selection operator above can keep the ascendant of better individuals.2.5. Crossover operatorIn crossover operator, offspring should inherit important factors of the parents. In our design of crossover operator, one-point crossover is adopted. And the crossover procedure is as follows:Step 1: Select two individual c 1 and c 2 from the whole individuals randomly.Step 2: Generate two crossover points at random.Step 3: Exchange the genes between the two crossover points for the two individual c 1 and c 2, and we can get two new individuals d 1 and d 2.For example, c 1 and c 2 are the two selected individuals.JIC email for contribution : submit@Journal of Information and Computing Science, 1 (2006) 1, pp 61-646362514632:1c43652153:2cThe crossover points generated randomly are 2 and 6. In order to get the new individuals, we copy the genes in c 1 before the 2nd gene and after the 6th gene into the same position of d 1, and copy the other genes of c 1 into the same position of d 2. In the same way, we copy the genes in c 2 before the 2nd gene and after the 6th gene into the same position of d 2, and copy the other genes of c 2 into the same position of d 1. Thus we obtain the whole new individuals d 1 and d 2. The process of crossover operation is shown in figure 1.62514632:1c62652152:1d 43652153:2c 43514633:2d CrossoverFigure 1: Example of crossover operation2.6. Mutation operatorIn our HGA, we adopt a double-point dynamic mutation operator, which is applied to change the values of the genes in the individuals. The mutation point and mutation value are generated randomly for every individual in each mutation operation. In each mutation operation, the mutation value is generated from 1 to N randomly.2.7. Outline of algorithmFrom the sections above, we can summarize the outline of our hybrid genetic algorithm, which is described as follows.Step 1: Generate the initial population G (t ), and set t ←1, the variable t represents the generation.Step 2: Pack the sequenced items into the bin sequence gained by NF procedure.Step 3: Evaluate the fitness of the individuals, and sort the individuals in the order of non-increasing value of fitness.Step 4: If E ≥T , terminate the algorithm and list the solutions obtained. The parameter E and T represent the fitness value and the given satisfying value of fitness (utility ratio of bin capacity), respectively. Otherwise continue.Step 5: Apply the selection operation to the sequenced individuals by the method of section 2.4.Step 6: Apply crossover operation to the individuals by the method of section 2.5.Step 7: Generate the mutation points and mutation values randomly for every individual. Apply the mutation operation to the individuals. Set t ←t+1, G (t ) ← G (t+1), and return to step 3. 3. Experimental resultsIn order to examine the performance of the hybrid genetic algorithm, an experiment is performed, and the experimental results are shown.For example, there are 20 sequenced items {I 1,I 2,…,I 20} and 4 kinds of bins {B 1,B 2,B 3,B 4}. The problem is how to pack the sequenced items into the bins so that the utility ratio of bin capacity is not less than 0.95. The sizes of the items S i and the capacities of the bins C i are shown as follows.41778135141761820i i i i i S ⎧⎪⎪⎨⎪⎪⎩≤≤≤≤=≤≤≤≤,121132153164i i i i i C ⎧⎪⎪⎨⎪⎪⎩===== In this experiment, the population size is 100, the crossover probability is 0.8, and the mutation probability is 0.02. The procedure based on HGA has run for 100 times, and every time satisfying solutions are obtained. The average genetic generation is 9.75 when the satisfying solutions are obtained. Three of the satisfying solutions obtained by hybrid genetic algorithm are shown in table 1.JIC email for subscription : info@64X. Jing, et al: A Hybrid Genetic Algorithm for Bin Packing Problem Based on Item SequencingTable 1: The experimental results for bin packing problemSolution 1 Solution 2 Solution 3Bin numberBin kind Item Bin kind Item Bin kind Item1 B1I1I2I3B4I1I2I3I4B2I1I2I32 B4I4I5I6I7B1I5I6I7B4I4I5I6I73 B3I8I9B3I8I9B3I8I94 B3I10I11B3I10I11B3I10I115 B3I12I13B3I12I13B3I12I136 B3I14I15I16B3I14I15I16B3I14I15I167 B1I17I18B1I17I18B1I17I188 B2I19I20B1I19I20B1I19I20Utility ratio0.9558 0.9643 0.9558As is shown in table 1, the utility ratios of the three solutions are all over 0.95. Obviously, the solution 2 is better than the other two. Although solution 1 and solution 3 have the same utility ratio of bin capacity, the bins used in each solution are not the same. This method can give us a chance to select a solution that we like from several satisfying solutions.4.ConclusionIn this paper, a hybrid genetic algorithm is proposed for solving one-dimensional bin packing problem based on item sequencing. For obtaining a better solution, the idea of simple genetic algorithm is used to search the solution of the bin sequence, and the idea of next fit algorithm is used to pack the sequenced items into the bin sequence obtained by simple genetic algorithm. Finally, the effectiveness of the hybrid genetic algorithm is convinced through computational results of an example. The method, based on the hybrid genetic algorithm for item sequencing based bin packing problem, can give us a chance to select the solutions that we like from several satisfying solutions, which is of great value in its application.5.References[1]M.R.Garey and D.S.Johnson, Computers and Intractability: A Guide to the Theory of NP-completeness,W.H.Freeman and Company, 1979.[2]J. H. Holland, Adaptation in Natural and Artificial Systems: An Introductory Analysis with Application to Biology,Control, and Artificial Intelligence, University of Michigan, 1975.[3]Erik D. Goodman, Alexander Y. Tetelbaum, and Victor M. Kureichik, A Genetic Algorithm Approach toCompaction, Bin Packing, and Nesting Problems, Michigan State University, July 12, 1994.[4]Hitoshi Iima, Tetsuya Yakawa, A New Design of Genetic Algorithm for Bin Packing, The 2003 Congress onEvolutionary Computation, 2(2003), 1044-1049.[5]Krzysztof Fleszar, Khalil S.Hindi, New Heuristics for One-dimensional Bin Packing, Computers & OperationsResearch 29(2002), 821-839.[6]Bruno Codenotti, Gianluca De Marco, Mauro Leoncini, Manuela Montangero and Massimo Santini,Approximation Algorithms for a Hierarchically Structured Bin Packing Problem, Information Processing Letters, 89(2004)5, 215-221.[7]Gregory Gutin, Tommy Jensen and Anders Yeo. Batched Bin Packing, Discrete Optimization, 2(2005)1, 71-82.[8] D.Hochbaum(ed.), Approximation Algorithms for NP-Hard Problems, PWS Publishing, Boston (1996), 46-93.JIC email for contribution: submit@。