Smooth constraint inversion technique in genetic algorithms and its application to surface wave study in the Tibetan Plateau

Smooth constraint is important in linear inversion, but it is difficult to apply directly to model parameters in genetic algorithms. If the model parameters are smoothed in iteration, the diversity of models will be greatly suppressed and all the models in population will tend to equal in a few iterations, so the optimal solution meeting requirement can not be obtained. In this paper, an indirect smooth constraint technique is introduced to genetic inversion. In this method, the new models produced in iteration are smoothed, then used as theoretical models in calculation of misfit function, but in process of iteration only the original models are used in order to keep the diversity of models. The technique is effective in inversion of surface wave and receiver function. Using this technique, we invert the phase velocity of Raleigh wave in the Tibetan Plateau, revealing the horizontal variation of S wave velocity structure near the center of the Tibetan Plateau. The results show that the S wave velocity in the north is relatively lower than that in the south. For most paths there is a lower velocity zone with 12-25 km thick at the depth of 15-40 km. The lower velocity zone in upper mantle is located below the depth of 100 km, and the thickness is usually 40-80 km, but for a few paths reach to 100 km thick. Among the area of Ando, Maqi and Ushu stations, there is an obvious lower velocity zone with the lowest velocity of 4.2-4.3 km/s at the depth of 90-230 km. Based on the S wave velocity structures of different paths and former data, we infer that the subduction of the Indian Plate is delimited nearby the Yarlung Zangbo suture zone.

Modal characteristics analysis for a flexible rotor with non-smooth constraint due to intermittent rub-impact

Intermittent rub-impact, during which the contact between rotor and stator is characterized by a ‘‘bouncing＂ or intermittent type of behavior, is one of the most common rubbing forms in rotating machinery. When the intermittent rub-impact occurs, the non-smooth constraint, which is the phenomenon that the system stiffness changes with respect to the state of contact and noncontact, will appear. The paper aims at discovering the possible effects of the non-smooth constraint on the flexible rotor＇s modal characteristics by theoretical and experimental methods. The qualitative description for non-smooth constraint is given for the intermittent rub-impact process, and the dynamic modeling for a rotor system with non-smooth constraint is carried out. Meanwhile, the analysis method is developed by Floquet theory and Hill＇s method to obtain the rotor＇s modal characteristics. The results reveal that the non-smooth constraint produced by the intermittent rubimpact will increase the modal frequencies and critical speeds of the rotor system significantly.Due to the time-varying features of the constraint stiffness, the modal frequencies for the intermittent rub-impact rotor present fluctuant changes with the increase of rotation speed, which is different from the general linear rotor system. The non-smooth constraint is possible to lead the rotor＇s instability, and the rotor＇s instable regions can be expanded significantly for the increase of average constraint stiffness, constraint amplitude and contact time ratio. Non-smooth constraint could also expand the resonance speed and resonance sideband of the rotor system, which sometimes results in amplitude jump phenomenon.

A new smoothing technique for mathematical programs with equilibrium constraints

A kind of mathematical programs with equilibrium constraints （MPEC） is studied. By using the idea of successive approximation, a smoothing nonlinear programming, which is equivalent to the MPEC problem, is proposed. Thereby, it is ensured that some classical optimization methods can be applied for the MPEC problem. In the end, two algorithm models are proposed with the detail analysis of the global convergence.

Numerical method for dynamics of multi-body systems with two-dimensional Coulomb dry friction and nonholonomic constraints

Based on the dynamical theory of multi-body systems with nonholonomic constraints and an algorithm for complementarity problems, a numerical method for the multi-body systems with two-dimensional Coulomb dry friction and nonholonomic constraints is presented. In particular, a wheeled multi-body system is considered. Here, the state transition of stick-slip between wheel and ground is transformed into a nonlinear complementarity problem （NCP）. An iterative algorithm for solving the NCP is then presented using an event-driven method. Dynamical equations of the multi-body system with holonomic and nonholonomic constraints are given using Routh equations and a con- straint stabilization method. Finally, an example is used to test the proposed numerical method. The results show some dynamical behaviors of the wheeled multi-body system and its constraint stabilization effects.

位姿约束下的双向扩展机械臂路径规划方法

针对机械臂路径规划方法存在的规划效率低、连杆通过性差、路径粗糙等问题,以渐进最优快速随机搜索树RRT*为基础,提出一种位姿约束下的双向扩展机械臂路径规划方法(PCO-BT-RRT*)。首先,设计目标偏置引导的双向扩展RRT*算法(BT-RRT*),改进了RRT*算法的初始化过程,将起始点和目标点分别作为两棵随机树的初始节点,并通过目标偏置策略引导其以一定概率相向生长,加快探索未知区域,在保证路径代价较低的同时提升路径生成速度。其次,提出一种位姿约束路径优化策略(PCO),采用机械臂运动学模型和碰撞检测规则共同约束新节点扩展过程,寻找机械臂可达空间内的避障路径;对生成路径剪枝剔冗,缩短可行路径长度,同时对消冗节点以迭代调整的方式进行平滑优化,提高路径生成质量。通过仿真实验分析,验证了所提方法在路径规划问题上的显著性成效;在自主研发的BIM信息融合下建筑砌筑系统进行真机避障测试,验证了该方法的实用性。

The Non-Smooth Problem of Wheel Motion

Planar motion of a non-deformable wheel under the action of non-ideal unilateral constraints is considered. The mathematical description of this phenomenon has a form of a non-smooth initial value problem. The non-smoothness of this problem means that its solution is determined by an absolutely continuous function having a discontinuous first derivative. For this reason, a collision problem describing abrupt changes of velocity has been formulated next to the equations of motion specifying the acceleration. The non-idealness of constraints means that the constraint reaction force includes also a component resulting from the friction between the wheel and the constraints. Differential equations specifying acceleration of the wheel making contact with the constraints and algebraic equations for determining the changes in the wheel’s velocity at the moment of collision have been formulated in the paper. The principal task in these formulations is to determine the reaction forces of the considered constraints. This task is specified by the relationships between acceleration and the constraint reaction force components. In the description of the collision, these relations refer to the post-collision velocities and reaction force impulses. For determining an approximate solution of the formulated wheel motion problem, an original numerical method and a computer program for wheel motion simulation have been developed. Selected results illustrating the changes in displacements and velocity have been presented.

A Fast Approach for Time Optimal and Smooth Trajectory Planning of Robot Manipulators

In this paper, a fast approach to generate time optimal and smooth trajectory has been developed and tested. Minimum time is critical for the productivity in industrial applications. Meanwhile, smooth trajectories based on cubic splines are desirable for their ability to limit vibrations and ensure the continuity of position, velocity and acceleration during the robot movement. The main feature of the approach is a satisfactory solution that can be obtained by a local modification process among each intermal between two consecutive via-points. An analytical formulation simplifies the approach to smooth trajectory and few,iterations are enough to determine the correct values. The approach can be applied in many robot manipulators which require high performance on time and smooth. The simulation and application of the approach on a palletizer robot are performed, and the experimental results provide evidence that the approach can realize the robot manipulators more efficiency and high smooth performance.

THE TOPOLOGICAL OPTIMIZATION FOR TRUSS STRUCTURES WITH STRESS CONSTRAINTS BASED ON THE EXIST-NULL COMBINED MODEL

A new exist-null combined model is proposed for the structural topology optimization. The model is applied to the topology optimization of the truss with stress constraints. Satisfactory computational result can be obtained with more rapid and more stable convergence as compared with the cross-sectional optimization. This work also shows that the presence of independent and continuous topological variable motivates the research of structural topology optimization.

复杂背景下基于SIFT算法的局部遮挡人脸识别

由于复杂的背景环境中存在大量噪声点、平滑点以及失真点,影响人脸识别效果,提出基于SIFT算法的局部遮挡人脸识别方法。考虑到噪声因素的影响建立尺度不变高斯函数,通过该函数描述原始图像的噪声干扰,查找平滑程度最高的像素点,以像素均值作为参考,对像素点及周围区域实行平滑约束。建立包含眼睛、鼻子、嘴巴以及耳朵的面部特征子集,与标准图像比对,采用最大期望算法提取重叠度最高的面部特征,作为人脸识别的参照标准。采用两个向量的内积衡量人脸图像与参考值之间的相似度,计算存在和不存在局部遮挡的特征权重,建立置信区间,代入面部特征权重实施置信对比,完成人脸识别。实验结果表明,所提方法的识别精准度较高,在灰度变化、噪声以及局部遮挡的情况下均能够保证识别精度。

基于峰值感知和多尺度约束的加权引导滤波器

引导图像滤波无法保留锐利边缘,导致平滑图像出现结构过度模糊问题,为此提出新的加权引导图像滤波器.利用峰值感知加权提取图像的边缘和结构信息,通过多尺度约束提高所提滤波器的鲁棒性.基于图像方差信息将滤波损失函数的正则化项系数改进为自适应式.边缘感知平滑、图像细节增强、纹理去除平滑以及图像去噪领域的应用实验结果表明,所提滤波器在视觉效果、峰值信噪比以及结构相似度上均优于参与对比的引导图像滤波器.与边缘感知平滑实验中次优滤波器的峰值信噪比和结构相似度相比,所提滤波器的峰值信噪比平均高2.62 dB,结构相似度平均高0.0286.

基于DIoU损失与平滑约束的结构化SVM目标跟踪方法

基于结构化SVM的目标跟踪因其优良的性能而受到了广泛的关注,但是现有方法存在损失函数不精确和模型漂移问题。针对这两个问题,首先提出基于DIoU损失与平滑约束的结构化SVM模型。该模型采用DIoU函数作为损失函数,利用t时刻超平面法向量w_(t)与t-1时刻超平面法向量w_(t-1)差值的L_(2)范数作为平滑约束。其次基于对偶坐标下降原理设计了该模型的求解算法。最后利用提出的基于DIoU损失与平滑约束的结构化SVM实现了一种多尺度目标跟踪方法。对所提出的目标跟踪方法在OTB100和VOT-ST2021数据集上进行了实验验证,实验结果表明:所提出的Scale-DCSSVM在OTB数据集上的跟踪成功率比DeepSRDCF高1.1个百分点,在VOT-ST2021上的EAO比E.T.Track高1.2个百分点。所提方法具有较优的性能。

重力三维密度光滑反演软件设计与实现

采用科学计算语言MATLAB开发一个基于光滑约束的重力三维密度反演软件,包含三维密度反演、重力场正演、数据分析和可视化等功能。测试结果表明,该软件能够反演得到合理的地质体密度分布,具有可靠性和实用性。

基于弹载图像的代价敏感与平滑约束结构化SVM目标跟踪方法

目标自寻的炮弹打击目标时,传统的目标跟踪方法易受混杂背景的影响,产生模型漂移从而导致跟踪失败。近年来,随着结构化SVM技术的发展,基于结构化SVM的目标跟踪方法能够有效解决复杂背景的问题。弹载成像条件下,在结构化SVM目标跟踪方法的基础上,增加代价敏感权重解决弹载图像中背景混杂所引起的正、负样本不平衡问题,同时利用t时刻超平面法向量w_(t)与t-1时刻超平面法向量w_(t-1)差值的L_(2)范数作为平滑约束抑制模型漂移问题。通过基于对偶坐标下降原理设计了模型的求解算法并实现一种多尺度目标跟踪方法Scale-CS_SSVM。在弹载数据集上进行实验验证,结果表明:与Scale-DLSSVM方法相比,Scale-CS_SSVM在跟踪精度和成功率上分别提高了5.5和5.0个百分点,达到了最优的性能。

基于机器视觉特征强化的降质图像动态目标检测方法

针对降质图像中的动态目标特征不明显,易出现检测偏差的问题,提出基于机器视觉特征强化的降质图像动态目标检测方法。采用帧间阈值法确定图像中含有动态目标的区域,通过平滑、动态目标强化和背景抑制处理,引入SE-ResNeXt模块,基于挤压-激励操作后提取动态目标的特征信息,计算动态目标与背景间的速度和光流矢量,在速度平滑性、方向平滑性和颜色分量约束条件下,检测得到动态目标的准确位置。结果表明,所提的方法在准确率、召回率和F 1分数的最优参数组合下,PSNR和SSIM平均提高了3.41%和0.16%,提升了降质图像整体质量,实现了对动态目标的精准检测与识别。

基于RRT算法的移动机器人安全光滑路径生成

在多障碍物复杂工厂环境中,针对快速探索随机树算法(RRT)生成的路径存在冗余点、贴近障碍物且存在锯齿状转折的问题,改进得到了安全-光滑RRT(Safe-SmoothRRT)路径规划算法。首先,引入目标偏置策略;其次,该算法利用融合目标点引力思想的新节点扩展方式以及改进的近邻点度量策略以减少树的盲目扩展,提高生长的目标性;随后,引入节点安全约束,将安全节点加入树中;改进路径简化方法,剔除冗余点的同时兼顾了安全性;最后通过B样条局部平滑来改善路径的平滑性。在MATLAB仿真实验中分别与标准RRT算法、自适应目标偏向性RRT算法和改进RRT算法相比,在平均路径长度方面最大下降了7.1%,在平均有效节点数方面最大下降了64.1%,且所得路径始终与障碍物保持一定的安全距离,结果表明改进算法有效提升了路径的光滑性和安全性。

基于双平滑函数秩近似和群稀疏的高光谱图像恢复模型

高光谱图像(HSI)具有良好的光谱识别能力,被广泛地应用于各种领域。然而,HSI在成像过程中易受到混合噪声的污染,会严重削弱后续任务的准确性,如何高质量地恢复HSI是需要解决的首要问题。目前,基于低秩先验和全变分正则化结合的HSI去噪方法取得了较好的性能,但这些方法一方面忽略了高强度条纹噪声在空间结构和光谱分布上的特征,使得噪声无法完全去除,另一方面没有考虑HSI差分图像低秩子空间的信息,不能挖掘潜在的局部空间光滑结构。为此,提出了一种基于双平滑函数秩近似和群稀疏的HSI恢复模型。首先,利用双平滑函数秩近似模型探索干净HSI和条纹噪声的低秩结构,去除结构化条纹噪声等高强度混合噪声。其次,将基于E3DTV的群稀疏正则化融入双平滑函数秩近似模型中,充分挖掘HSI差分图像的稀疏先验信息,进一步提升图像在空间恢复和光谱特征保留方面的性能。最后,设计了交替方向乘子法(ADMM)求解所提出的BSRAGS模型。仿真和真实数据实验均表明,所提模型能够有效提高图像恢复质量。

一种考虑关节跃度约束的实时刀具路径光顺算法

针对机器人加工过程中关节跃度超过约束的问题,提出一种考虑机器人关节跃度约束的刀具路径光顺算法。采用有限冲激响应滤波器插值算法对五轴刀具路径进行光顺处理,一步完成了路径光顺和速度规划。建立了关节与刀具之间的速度、加速度和跃度关系,通过调整时间常数保证了关节跃度不超过约束。考虑机器人关节加速度约束后,该算法计算出的时间常数更小。仿真和实验结果表明该算法能减小关节的跟踪误差,提高机器人的工作稳定性。

融合目标导向和跳跃点改进A算法的煤矿巡检机器人路径规划

针对煤矿巡检机器人采用A算法规划路径效率低、安全性差等问题,提出了一种融合目标导向和跳跃点改进A算法的煤矿巡检机器人路径规划方法。首先,在传统A算法中引入碰撞约束机制,改进节点搜索条件,使得获得的节点与周围障碍物节点保持一定的安全距离;其次,通过融合目标导向和跳跃点优化A算法,剔除路径搜索过程中无关扩展节点的搜索,以提升路径搜索效率;最后,利用分层平滑优化策略消除路径中的冗余点和转折点,减少路径长度并提高路径平滑性。仿真结果表明,该算法能够有效提升煤矿巡检机器人的工作效率和安全性。

分层平滑优化A*引导DWA用于机器人路径规划

针对A∗算法存在的搜索效率低,路径平滑性和安全性差,以及DWA融合全局路径规划算法实时寻路效率低等问题,提出了一种分层平滑优化A∗引导DWA(HSA∗-G-DWA)的移动机器人路径规划方法。首先,在A∗算法的代价函数中引入双动态加权因子并构建碰撞约束函数剔除路径搜索过程中无关扩展节点的搜索,以提升路径搜索的效率和安全性。其次,利用分层平滑优化策略消除路径中的冗余点和转折点,减少路径点数量和路径长度。之后,通过无障碍约束直线与有障碍约束圆弧插补分段优化生成初始全局路径,保证路径的安全性与平滑性。然后,若移动机器人跟踪全局路径过程中面临未知障碍物则利用全局路径引导DWA生成避障与返回全局路径的局部动态修正路径,减少了实时计算量。最后,仿真实验结果表明,静态环境下HSA∗-G-DWA算法路径搜索时间和路径点数较A∗算法分别平均减少了88.43%和86%,路径的平滑性和安全性更好;未知环境下HSA∗-G-DWA算法可以实时避开环境中出现的未知障碍物,路径长度较DWA算法、Dijkstra算法、RRT算法和现有融合算法分别平均减少了25.78%、18.65%、30.48%和14.59%,路径搜索时间较现有融合算法平均减少了67.39%。

用改进的光滑约束最小二乘正交分解法实现电阻率三维反演

对三维电阻率反演问题进行了深入研究,提供了一种利用地表观测数据实现三维反演的实用算法.该方法应用有限差分求正演解,并通过对粗糙度矩阵元素进行适当改进,使之适用于各种情况下粗糙度矩阵的求取,进而建立在模型的总粗糙度极小条件下的反演方程.对反演方程采用收敛速度快且稳定的最小二乘正交分解(LSQR)法进行迭代求解,在迭代求解过程中只需利用偏导数矩阵和其转置矩阵乘以一个向量的结果,回避了直接求偏导数矩阵的繁琐计算,节省了内存,加快了反演的计算速度.不同的计算实例表明上述方法是求解大规模三维电阻率反演问题的有效方法.