How to Use Driving Simulators Properly: Impacts of Human Sensory and Perceptual Capabilities on Visual Fidelity

Fidelity has been a critical concern of researchers throughout the history of driving simulation. Understanding the limits of a driving simulation system is a prerequisite for conducting valid driving simulator studies. This paper proposes a novel and interdisciplinary methodology to ensure validity of studies using driving simulators (primarily for traffic control devices and other object detection tasks) based on the visual limits of human sensory and perceptual capabilities, and the characteristics of raster graphics. This methodology decomposes the perceptual issues of a stimulus into perceptual issues of different visual properties like luminance, hue, or text of the stimulus. By systematically analyzing the mechanism of human vision in driving simulators, the perceptual principle is proposed to ensure perceivable visual details in human-in-the-loop driving simulation systems. Additionally, the graphic principle is proposed to ensure perceivable features of a target object in the virtual driving environment. Both principles quantify the minimum requirements of visual fidelity with two measurements: angular resolution and matrix dimensions. The enriched results from existing pertinent studies are analyzed and organized to yield support of both principles. This research focuses on the minimum requirements for four factors; namely the visual acuity of drivers, the specifications of display systems, the configurations of graphics systems, and the design of virtual scenarios, as well as the relationship among all these factors to assess the visual fidelity in driving simulation systems. Within the realm of human perception, this work can provide criteria for proper design, calibration, and usage of driving simulators.     

基于奇异值分解的混沌时间序列Volterra预测

提出一种用于混沌时间序列预测的奇异值分解Volterra滤波器.在Volterra滤波系数计算过程中,采用奇异值分解的方法得到线性方程组的最小二乘解.4种混沌序列的预测实验表明:该滤波器对混沌流的预测性能远优于NLMS自适应Volterra滤波器,前者的一步预测相对误差比后者小3~4个数量级.     

基于奇异值分解的灰色模型参数估计

提出了一种新的基于奇异值分解的GM(1,1),GM(1,N)等灰色模型参数估计方法,该方法在进行奇异值分解时,借助2个辅助矩阵将奇异值分解矩阵分解为多个矩阵之积.研究了该估计方法下对原始序列进行数乘变换后参数的变化特征.研究表明该方法充分利用了奇异值分解的优点,回避了由于最小二乘法给灰色模型带来病态性的缺点,提高了模型的稳定性,同时还减少了计算量.     

环境激励下钢筋混凝土拱桥模态参数识别

为正确选择模态参数识别方法,对某钢筋混凝土拱桥进行了环境振动试验,采用频域、时域和时频分析方法——峰值拾取法(PP)、随机子空间识别(SSI)以及基于经验模态分解(EMD)和随机减量技术(RDT)的方法——识别其模态参数;比较了3种模态参数识别方法的特点和识别结果,并与有限元法计算结果进行了比较.研究结果表明:峰值拾取法的识别速度快,识别的频率较可靠,但识别过程需要较多人工干预;随机子空间识别的理论体系完备,适合程序实现,识别过程能较好地抵消测试过程中噪声等的影响;由于模态混叠的影响,EMD-RDT识别结果具有一定随机性.     

Formation of nanometer coherent structures during spinodal decomposition and ordering coexistence phase transformation in Fe-24Al alloys

The nanometer coherent structure evolution of spinodal decomposition and ordering coexistence phase transformation in Fe-24Al alloys is investigated by the microscopic phase field kinetic model. The results show that the concentration and long-range order parameters all continuously change towards to their equilibrium values during phase transformation. With the increase of elastic interaction energy, the anisotropy along [01] or [10] elastic soft direction is more obvious and the time reaching equilibrium state is also shortened. According to the results, the formation of nanometer coherent structures during phase transformation is composed of the initial decreasing stage of order degree stage, the incubation stage, the continuous increasing stage of concentration order parameter and long-range order parameter, and the later stable stage. The spinodal decomposition and ordering is interaction; the initial ordering stage is a necessary condition of the coexistence phase transformation. The nanometer coherent structures are not found to grow during the whole phase transformation. The simulation results are in accordance with the results in experiment obtained by the aging treatment in Fe-24Al alloys.     

隧道中地铁车辆内部噪声分析

采用Artemis测试分析系统对隧道内运行的大连厂地铁车辆进行噪声测试,在地铁车辆内选择了六个测试点,通过对测试数据的分析、讨论.并对测试数据进行了分析,结果表明：地铁车辆运行时,车辆内噪声的最主要的噪声源是轮轨噪声.噪声级随着地铁车辆的速度的增加而增加.主频带一般都在315～5 000 Hz之间.低频率的声压级很小.研究地铁车辆内的噪声特性只需研究中高频声压级.     

HTN规划中面向多计划生成的顺序任务分解算法

顺序任务分解算法（OTD）是层次任务网规划（HTN）中的一种高效求解算法．由于算法中的计划生成采用一次性回溯机制,每次求解过程只能产生一个可行计划．文中提出了一种能够快速生成多个可行计划的回溯算法．该算法采用分段回溯的计划生成机制,充分利用了求解过程中生成的局部解序列,从而能够一次性地快速生成多个可行计划,为寻求优化的计划和进行计划的评估提供更为有效、灵活的支持．     

基于时间-费用约束的虚拟物流企业项目分解研究

考虑时间和费用约束条件下的虚拟物流企业项目分解.找出虚拟物流网络中的最短时延路,若该路的时延大于时间约束,则虚拟物流企业盟主认为不能在规定时间内完成该虚拟物流项目,拒绝该物流项目;反之,则从所有满足时间约束的路中挑出一条最小费用的路,通过此路进行分解,若时间和费用都满足虚拟物流企业盟主的期望,则接受该虚拟物流项目.并给出一个算例.     

Network Decomposition and Maximum Independent Set Part Ⅰ:Theoretic Basis

IntroductionItisofsignificanttheoreticalmeaningtofindthemaximumindependentsetwithinanetwork .Thisisnotonlyrelatedtoclique ,nodecovering ,coloringandmatchingissues ,butalsoassociatedwithmanycombinatorialoptimization problems .In practicalapplications ,find…     

船舶机械振动源激励力的间接估计工程方法

船舶机械设备输出到船体上的激励力对船舶结构振动预报有着重要意义,文章讨论了两种间接估算激励力的方法,目的是希望在工程现场应用这些工程方法取代激励力的直接测量.对刚性安装的机器推荐采用最小二乘方法,该方法是基于支撑结构的加速度导纳特性进行估算.对弹性安装的机器而言,推荐采用基于隔振器阻抗特性的激励力间接估算方法.力的估算是逆问题,文中采用奇异值分解技术来处理会遇到的病态问题,从而尽量降低估算误差.文中还对估算的准确度、估算质量评价等进行了讨论.条件数的倒数在最小二乘法响应点的选取时可作为优选的先验判据,归一化残差可作为估算质量的后验判据.实验证明了间接估算方法是确定船舶机械激励力的恰当选择,实验结果的最小二乘法相对误差为3分贝.测试中的信噪比被认为在力估算过程中有重要影响,低信噪比很可能是实际测试时的首要误差来源.该文也可用于指导船舶振动源特性描述的研究.