元胞遗传算法的全局优化性能仿真分析

元胞遗传算法将遗传操作限制在邻域内进行,减缓了优势个体在群体中的扩散速度,具有更好的全局探索能力,在求解复杂优化问题中显示出优越性.与传统遗传算法对比,以选择压力作为分析手段,对元胞遗传算法进行定性分析.通过求解具有不同特征的函数,分析进化过程群体多样性变化.从进化过程群体分布图,直观得出元胞遗传算法具有较好的维持群体多样性能力;统计结果表明,元胞遗传算法能极大提高全局收敛率,并且求解稳定性更好.     

基于选择压力的全局优化元胞遗传算法分析

元胞遗传算法将遗传操作限制在邻域内进行,减缓了优势个体在群体中的扩散速度,具有更好的全局收敛性,在求解复杂优化问题中显示出优越性。与传统遗传算法对比,以选择压力作为分析手段,对元胞遗传算法进行定性分析。通过求解具有不同特征的函数,分析进化过程群体多样性变化,从进化过程群体分布图,直观得出元胞遗传算法具有较好的维持群体多样性能力;从计算的统计结果,得出元胞遗传算法能极大提高全局收敛率,并且求解稳定性更好。     

基于分布式遗传算法的移动Agent迁移策略

迁移策略是移动Agent（Mobile Agent，MA）的核心技术之一，MA的效率很大程度上取决于迁移策略的优化。本文提出了一种改进的分布式遗传算法（EDGA），用于对多约束条件下MA迁移策略最优问题进行求解。EDGA将分布式遗传算法和Cascade模型相结合，在迁移算子部分设计一个中心监控器，观察每个子种群的进化，并对迁移个体的选择以及相应子种群的大小做出调整，使进化能力好的子种群得到更大的空间来搜索最优值。实验结果表明：本文所提出的EDGA算法在求解速度和质量上取得了较大的改善。     

多约束QoS移动IP最优路由的求解

提出了一种改进的分布式遗传算法（EDGA）和链路编码方法用于解决多约束服务质量（QoS）移动IP最优路由选择问题。将分布式遗传算法和Cascade模型相结合，在迁移算子部分设计一个中心监控器，观察每个子种群的进化，并对迁移个体的选择以及相应子种群的大小做出调整，使进化能力好的子种群得到更大的空间来搜索最优值。实验结果表明，该EDGA算法在求解速度和质量上取得了较大的改善。     

基于精英选择和个体迁移的多目标遗传算法

提出基于遗传算法求解多目标优化问题的方法,将多目标问题分解成多个单目标优化问题,用遗传算法分别在每个单目标种群中并行搜索.在进化过程中的每一代,采用精英选择和个体迁移策略加快多个目标的并行搜索,提出了控制Pareto最优解数量并保持个体多样性的有限精度法,同时还提出了多目标遗传算法的终止条件.数值实验说明所提出的算法能较快地找到一组分布广泛且均匀的Pareto最优解.     

基于元胞多目标遗传算法的蜗杆优化设计

为研究蜗杆传动的多目标优化问题,提出一种自适应差分进化的元胞多目标遗传算法。该算法针对元胞遗传算法的特点,对基本的差分进化策略进行改进,得到一种参数自适应控制策略。将该算法与目前性能优异的4种多目标进化算法在三目标的基准测试函数进行对比实验,结果表明所提算法相对于其他算法具有明显的优势,能够在保证良好收敛性的同时,使获得的Pareto前端分布性更加均匀,覆盖范围更广;工程实例求解结果也表明了算法的工程可行性。     

基于改进元胞多目标遗传算法的机床主轴优化

为了进一步提高元胞遗传算法在求解多目标优化问题时的收敛性和分布性。在多目标元胞遗传算法的基础上,引入了三维空间元胞,提出了三维元胞多目标遗传算法。采用多目标基准测试函数对该算法进行了测试,并将其与目前比较流行的几种多目标遗传算法进行对比。结果表明,此种算法在收敛性和分布性上取得了更好的效果。采用以上这几种算法分别对机床主轴多目标优化问题进行了求解,相比其他几种算法,改进的多目标元胞遗传算法得到了更优的结果,说明了改进的算法在求解此问题时行之有效。     

一种基于混沌迁移的伪并行遗传算法及其应用

为了解决遗传算法寻优过程中的早熟收敛问题 ,本文提出了一种基于混沌迁移策略的伪并行遗传算法 ,该算法针对实时性要求不高的优化问题采用串行的算法结构实现分解型并行遗传算法的“独立进化、信息交换”思想 .在并行进化的个体异步迁移过程中 ,引入了混沌迁移序列引导个体迁移过程 ,利用其遍历性和随机性 ,保证了子种群之间能够进行充分高效的信息交换 .仿真研究和在库存优化方面的应用研究表明 ,这种算法具有很强的全局搜索能力 ,寻优效率高 ,有效克服了标准遗传算法的早熟收敛问题 .     

一种动态邻居元胞遗传算法

研究一种改进的元胞遗传算法。将遗传算法中的个体适应度和元胞自动机中的邻居定义做了结合,提出基于元胞间距离以及元胞个体适应度的"影响力算子",并作为算子中心元胞判断邻居的依据,从而形成改进算法,并对改进算法的基本性能的进行了两组定量分析,一是影响力算子对选择压和多样化损失的控制,另一部分是将算法与改良后传统元胞遗传算法做了对比测试。结果显示,即便使用最朴素的影响力算子而且不采用其它优化手段的情况下,算法依然能对选择压和多样化损失进行有效地控制,并且相较于使用了最优个体保持和小范围竞争择优的传统元胞遗传算法收敛率提高了约10%。     

基于记忆库拉马克进化算法的作业车间调度

多种群遗传算法相比遗传算法在性能上能够有所提高,但对具有较多局部最优解的作业车间调度问题,多种群遗传算法仍然难以改善易陷入局部最优解和局部搜索能力差的缺点.因此,提出了一种求解作业车间调度问题的新算法MGA-MBL(multi-population genetic algorithm based on memory-base and Lamarckian evolution for job shop scheduling problem).MGA-MBL在多种群遗传算法的基础上通过引入记忆库策略,不但使子种群间的个体可以进行信息交换,而且有利于保持整个种群的多样性;通过构造基于拉马克进化机制的局部搜索算子来提高多种群遗传算法中子种群进化的局部搜索能力.由于MGA-MBL采用了全局寻优能力较强的模拟退火算法对记忆库中的个体进行优化,从而缓解了多种群遗传算法易陷入局部最优解的问题,并提高了算法求解作业车间调度问题的性能.对著名的benchmark数据进行测试,实验结果证实了MGA-MBL在求解作业车间调度问题上的有效性.     

求解广义旅行商问题的混合染色体遗传算法

提出了针对广义旅行商问题（GTSP）的混合染色体遗传算法（HCGA）。目前,广义染色体遗传算法（GCGA）是求解GTSP问题的最好方法,但这种方法在编码设计上存在不足,使得算法全局搜索能力较差。在GCGA算法基础上,设计了二进制和整数混合编码的染色体,并更新了交叉和变异算子设计,得到改进算法HCGA。理论和实验结果都说明：HCGA比GCGA等多种算法具有更强的全局搜索能力。     

启发式列生成算法求解带恶化效应的同构并行机调度问题

针对实际生产中广泛存在的一类带恶化效应的同构并行机调度问题,以最小化最大完工时间为优化目标,构建该问题的整数规划模型,并提出一种启发式列生成算法(HCGA)进行求解.在HCGA中,首先,利用Dantzig-Wolfe分解方法,将原问题分解为一个主问题(MP)和多个子问题;然后,设计启发式算法获得初始列,其中每列为一台机器上的一个调度方案,基于初始列构建限制主问题(RMP)模型;接着,设计快速有效的动态规划算法求解子问题,以得到需添加至RMP的列集,同时,考虑传统列生成算法收敛速度较慢,设计一系列方法来加速列生成过程;最后,基于所获取的MP线性松弛解,设计深潜启发式算法确定原问题的整数解.HCGA与商用求解器GUROBI的对比实验结果表明,HCGA可在较短时间内获得更优的解.     

Wireless visual sensor networks for smart city applications:A relevance-based approach for multiple sinks mobility

Nowadays, the major cities of the world have to solve problems that were unthinkable in past decades. Due to the population growing rate, new issues are still arising, but technology can be used to address such issues and improve life quality in big cities. In that scenario, surveillance is a highly desired service and most governments are already using different types of devices to provide high levels of security. Wireless Visual Sensor Networks (WVSN) can be used to monitor every part of a city without the cost of running cables all over it. However, there must be an efficient way to gather all information collected by the sensors and cameras, with reduced energy consumption and average latency. This work proposes a new algorithm to position multiple mobile sinks in WVSN deployed along roads and streets. A relevance-based approach was designed to position sinks closer to source nodes with higher sensing relevance, since they are expected to transmit more data packets. The proposed algorithm can detect forbidden and disconnected zones, making sure sinks will be positioned in permitted areas, which makes this approach very suitable for realistic smart city applications.     

A smart healthcare reward model for resource allocation in smart city

Today, cities face many significant challenges, and the smart city concept is a promising means to address typical traditional city problems. The wireless e-health technologies is an evolving topic in the area of telemedicine nowadays. Mobile telecommunication and the use of multimedia technologies are the core of providing better access to healthcare personnel on the move. These technologies provide equal access to medical information and expert care leading to a better and a more efficient use of resources. Mobile and Fog computing technologies can also cope with many challenges in smart healthcare resources of mobility, scalability, efficiency, and reliability. Optimal healthcare systems are particularly critical in cities, due to the highly concentrated populations. This high population increases the potential for harm and damage in the case of negligence or improper treatment. This can lead to infections and disease outbreaks, which could become epidemic situations and require containment, which is very costly. Motivated by the need for better usage and management of healthcare resources, which is crucial for reliable healthcare delivery, this paper introduces a model that can provide improved delivery and utilization of resources. The quality reward-based model was developed to study and react to the satisfaction factors of healthcare systems, and proposes an optimization-based algorithm called the Maximum Reward Algorithm (MRA), that enhances the use and delivery of healthcare resources. The algorithm has been tested with multiple experiments and simulations, and has proved that it can provide reliability, efficiency and achieves 50.1% to 77.2% performance improvement.

     

求解TSP问题的自适应邻域遗传算法

提出结合自适应邻域法与遗传算法来求解TSP问题。在自适应邻域法中,从某个城市出发,下一城市不一定是其最近城市,而是在比其最近城市稍远的邻域范围进行动态随机选取。在求解TSP时,采用自适应邻域法对种群初始化,然后采用选择、交叉、变异进行迭代,在选择中仅保留父代90％的样本,剩下的采用自适应邻域法产生新样本进行补充。仿真实验结果表明所提算法与其他算法相比具有竞争能力。     

Census from Heaven: An estimate of the global human population using night-time satellite imagery

Night-time satellite imagery provided by the Defense Meteorological Satellite Program's Operational Linescan System (DMSP OLS) is evaluated as a means of estimating the population of all the cities of the world based on their areal extent in the image. A global night-time image product was registered to a dataset of 2000 known city locations with known populations. A relationship between areal extent and city population discovered by Tobler and Nordbeck is identified on a nation by nation basis to estimate the population of the 22 920 urban clusters that exist in the night-time satellite image. The relationship between city population and city areal extent was derived from 1597 city point locations with known population that landed in a 'lit' area of the image. Due to conurbation, these 1597 cities resulted in only 1383 points of analysis for performing regression. When several cities fell into one 'lit' area their populations were summed. The results of this analysis allow for an estimate of the urban population of every nation of the world. By using the known percent of population in urban areas for every nation a total national population was also estimated. The sum of these estimates is a total estimate of the global human population, which in this case was 6.3 billion. This is fairly close to the generally accepted contemporaneous (1997) estimate of the global population which stood at approximately 5.9 billion.     

Hybrid control systems and the dynamic traveling salesman problem

A new approximate algorithm for solving the dynamic traveling salesman problem (DTSP) is proposed; the traveling salesman starting from the base city visits megapoleis and cities inside megapoleis and comes back to the base city. A specific feature of this variant of DTSP is the movement of cities inside megapoleis in time. To solve this problem, a general solution theory for hybrid (complicated) systems with “combinatorial” and “continuous” path segments is developed. The general theory is based on the sufficient optimality conditions known in the theory of optimal control.     

Has Dongying developed to a ghost city? - Evidence from multi-temporal population estimation based on VHR remote sensing and census counts

With ongoing growth and continuous development of cities, the world is turning into an urban society. In this context, urbanization, population growth and migration towards urban areas are global trends. These processes are highly dynamic especially in China, with highest rates of urbanization worldwide. In contrast to these well-known trends, a recently emerging and rarely studied side effect is “ghost cities”, which became a notable phenomenon in China recently. A ghost city is commonly defined as a new urban development that is running at severe undercapacity with respect to population and businesses and where the availability of housing and public infrastructure significantly exceeds the practical demand. Against this background, this study presents a framework based on remote sensing for the assessment of the presence or absence of the ghost city phenomenon in a typical highly dynamic Chinese city. For this purpose, remote sensing data with very-high resolution (VHR) are employed for establishment of a 4d functional city model. Subsequently population capacity estimates are based on a statistical approach. The components of the functional 4d city model, i.e. the multi-temporal building model and the classification of building types associated with residential and non-residential function returned very high accuracies with κ of 0.73 and 0.89, respectively. The number of floors was estimated with coefficient of determination of 0.91. Compared to the numbers of official census counts, the multi-temporal population capacity estimation revealed a considerable mismatch of available living space based on VHR remote sensing data and actual population counts. According to the conceptual framework of this study, this disagreement indicates a high likelihood and significant evidence for the emergence and presence of the ghost city phenomenon for the city of Dongying. In addition, a detailed spatial assessment was conducted in terms of an index comparing the dynamics of residential developments and population numbers to provide an impression of specific regions of the urban area which are most likely to suffer from the ghost city phenomenon.     

Agglomeration centrality to examine spatial scaling law in cities

Cities grow with processes resulting in the agglomeration of population and socioeconomic activities. At any stage of urban development, a city possesses a spatial structure of urban fabric to facilitate its functions. In this study, we find that the agglomeration characteristics of a city are significant for the identification and understanding of urban spatial structure. We propose a new centrality measure, termed agglomeration centrality, to examine the centrality of a node based on its relative level of clumpiness in an urban road network. We show a simple method to computationally define the agglomeration centrality based on Voronoi tessellations of all nodes in the network. On this basis, we investigate the spatial distribution of agglomeration centrality of urban road networks. Experiments on a set of randomly selected cities show that the agglomeration distributions of all these urban street networks present a remarkable double-power-law phenomenon. The distribution of the agglomeration centrality measures shows an invariant spatial scaling rule for cities. The regions with the highest centralities are always found to be urban centers. In comparison, this cannot be achieved by any current centrality measures. Finally, in an effort to explore the potential additional applicability of the new centrality measure to aid our understanding of spatial structures of other urban demographic or socioeconomic attributes, the agglomeration centrality was combined with the population data to create a population agglomeration index. The distribution of this index indicates a single-power-law property. In summary, this study develops a new centrality measure and demonstrates its effectiveness for understanding the spatial structures and spatial evolution of cities.