基于改进神经网络的多AUV全覆盖路径规划

针对三维环境下的多自主水下机器人(Autonomous Underwater Vehicle,AUV)全覆盖路径规划问题,提出一种基于改进神经网络—Glasius生物启发神经网络（Glasius Bio-inspired Neural Network,GBNN）的全覆盖路径规划算法。对AUV的水下工作环境构建离散的三维栅格地图;根据栅格地图,建立相对应的三维GBNN模型;根据GBNN活性值的动态变化,AUV规划各自的搜索路径,对水下任务区域进行全覆盖搜索。仿真结果表示,多AUV可以协同完成覆盖搜索任务,能够自动避开各类静态和动态的障碍物,自动逃离路径的死锁区。     

一种激光驾束制导武器半实物仿真系统构建及试验方法

激光驾束制导武器半实物仿真必须借助二维平动系统,但由于二维平动系统受限于技术指标不高,无法满足弹目相对运动模拟速度和加速度要求较高的武器系统开展仿真试验的弊端,提出一种新的激光驾束制导武器半实物仿真系统构建及试验方法。采用“二轴转台+负载支架”的仿真系统构建方式,利用二轴转台的转动角运动代替二维平动系统的线运动来实现导弹在激光信息场中弹目相对运动和位置变化的模拟,很好地解决了传统方法存在的种种弊端,试验结果进一步验证了该方法具有很好的仿真效果。     

基于树型空间的三维树木模型匹配

为解决三维树木模型的检索具有正确率低、检索效率低、描述不匹配等问题,提出一种基于树型空间的三维树木模型匹配方法。该方法通过构建三维树木模型数据集的树型空间,辅以主轴匹配,轮廓匹配,枝干匹配等方式,经过树型空间中欧氏距离和测地线度量差异值,逐步匹配到相应的三维树木模型,该方法分别在通用模型数据集和自我设计的模型数据集进行了匹配性实现,结果表明可以较大幅度提升树木模型匹配的准确率,利于三维树木模型的大规模检索和重用。     

基于泊位坐标的无人船自主靠泊仿真研究

针对神经网络自主靠泊控制器只能完成特定港口的靠泊,而无法拓展到其他没有训练数据的泊位这一问题,提出一种以泊位顶点为原点,岸线为y轴的坐标系(泊位坐标系)。采用泊位坐标系中相对位置来训练神经网络控制器,在V.Dragon-5000航海模拟器中选取集装箱船“银河号”,在大连港进行靠泊训练后,成功将靠泊控制器拓展到没有训练数据的深圳蛇口港与新加波樟宜港。仿真验证了采用新建立的坐标系可以拓展神经网络自主靠泊控制器适用范围,减小训练数据所需的成本。     

基于博弈论的化工园区合作巡逻策略研究

化工园区的生产活动对周围的大气环境和居民健康构成了巨大威胁,需要园区管理团队对化学生产过程和气体排放进行严格监管以保证空气质量,对各化工厂进行巡逻是一种有效的方法。基于博弈论研究多辆巡逻车的合作巡逻策略,提出了贪婪部署算法确定巡逻车的初始部署,采取静态分区的合作方法将化工园区划分成多个小区域,在化工厂和巡逻车之间构建巡逻博弈模型,通过MultiLPs算法计算得到博弈的Stackelberg均衡解。选取Antwerp化工园区进行案例研究,结果表明巡逻车可以将收益提高8.08%~35.24%。     

GROUP SIZE AND PARALLELISM EFFECTS IN TASKS WITHHETEROGENEOUS LEVELS OF DIFFICULTY：A STOCHASTIC ORDER APPROACH

We consider tasks where in order to perform them it is sufficient that one member of a group will know how to do it. We are interested in the effect of task difficulty, and variability of that difficulty, on group performance, and in particular on the marginal contribution of an additional number to the performance of groups of different size. We explore the implications of various stochastic orders over task difficulty and variability. Some intuitive conjectures are shown to be false.     

GROUP-BUYING ONLINE AUCTION AND OPTIMAL INVENTORY POLICY IN UNCERTAIN MARKET

In this paper we consider a group-buying online auction (GBA) model for a monopolistic manufacturer selling novel products in the uncertain market. Firstly, we introduce the bidder's dominant strategy, after which we optimize the GBA price curve and the production volume together. Finally, we compare the GBA with the traditional posted pricing mechanism and find that the GBA is highly probable to be advantageous over the posted pricing mechanism in some appropriate market environments.     

AUTOMATIC DECISION-MAKING： NEURAL FIRING AND RESPONSE

1. Introduction This is the fourth exposition on theAnalytic Hierarchy Process. It is true thatmany decisions we make subconsciously ariseout of our physical and biological needs andderive from the pressure to survive physically.We also make subconsciou…     

THE PERIODIC CAPACITATED ARC ROUTING PROBLEM LINEAR PROGRAMMING MODEL, METAHEURISTIC AND LOWER BOUNDS

1. Introduction The Capacitated Arc Routing Problem(CARP) is defined on an undirected network inwhich a fleet of identical vehicles with limitedcapacity is based at a depot node. Each edge hasa non-negative traversal cost and can betraversed any number…     

STUDIES ON HORIZONTAL COMPETITION AMONG HOMOGENOUS RETAILERS THROUGH AGENT-BASED SIMULATION

1. Introduction A supply chain is a network composed osuppliers, manufacturers, retailers, etc., whichcooperate to offer a kind of goods or servicesThese supply chain members cooperate andcompete with each other to maximize theiown profits. For example,…