基于BP神经网络的条纹级数插值方法

数据采集是光测力学数据处理自动化的关键和难点，特别是对条纹稀疏和低级数区．针对上述困难，本文利用ＢＰ神经网络能对函数进行逼近的特征，对条纹插值进行了探讨；经研究改进后的算法，其收敛速度明显提高．作为例子，给出了一幅光弹性等差线条纹图的插值实验，获得了令人满意的结果．     

密集模态挠性结构模型多变量频域辨识和控制

密集模态挠性结构的模态不仅阻尼小而且耦合程度高,给模型辨识带来了很大困难.频域辨识是获取空间挠性结构模型的一个有效方法,但是目前频域辨识要么模型结构定义有缺陷,要么是集中在单变量情形,不适合作大型空间挠性结构多变量模型的辨识.本文提出了多变量极大似然频域辨识,给出了其模型结构定义、算法推导及实现,并将其应用于H型密集模态挠性板的模型辨识,根据辨识结果,设计主动控制律,实现了对密集模态挠性板振动的有效抑制,表明了辨识算法及主动控制的可行性.     

适用于SR-CT技术的新算法

SR-CT(Synchrotron Radiation Computed Tomography)技术中有多种重建算法,其典型算法有滤波反投影算法和迭代算法,这两类算法各有其特点。综合考虑重建结果的质量和重建运算的时间,本文提出了一种新的算法混合算法。通过不同算法重建图像的分析比较,结果表明混合算法包含了前两类算法的优点,而且避免了它们的缺点,是一种行之有效的算法。同时对混合算法中的一些重要参数如初始解系数、迭代步长也进行了详细讨论,并给出了这些参数对重建图像质量的影响关系。     

用并行CG_DY算法进行圆柱体压力场测定

给出了由Dai和Yuan提出的非线性共轭梯度算法(CG_DY)的并行实现,并应用并行CG_DY算法在高性能计算机上求解圆柱体压力场测定的问题,比较了曙光2000和深腾6800的性能。数值计算给出了满意的结果,发现并行CG_DY算法在最优化问题的设计中具有很好的性质,并显示其具有很好的可扩展性。     

线性时不变系统集员辨识的区间算法

在不确定但有界(UBB)噪声假设下，提出一种针对线性时不变系统参数集员辨识的区间算法. 借助区间数学，寻求与观测数据和噪声相容的参数的最小超长方体(或区间向量)，推导了递 推列式，并分析了算法的收敛性. 此算法不仅可以给出参数估计值，还可以给出参数的不确 定性界限. 通过数值算例，将此算法与Fogel椭球算法和最小二乘算法进行了比较，显示 了其计算量小和精度高的优点。     

混凝土的一种标量损伤弹塑性本构模型

荷载作用下材料性能的劣化是混凝土结构的微观损伤机理,其宏观表现为结构刚度的折减和承载力的降低。论文推导了基于不可逆热力学过程的弹塑性标量损伤本构,给出时间离散的屈服准则。采用基于向后Euler法的应力更新算法——两步图形返回的最近点投影法,推导了满足迭代结果收敛假设的塑性参数及算法刚度张量,给出了空间梁单元本构积分算法的Jacobi矩阵。将模型用于混凝土简支梁的承载力试验模拟,与计算数据对比表明了模型和算法的合理性和有效性。     

基于离散正则化的实验载荷谱重构与推演算法

针对截齿截割煤岩随机过程中特征值提取较困难的问题,根据有限的截齿平面截割煤岩实验载荷谱,确定载荷谱的重构算法和波形特征,推演出了截齿截割煤岩的载荷谱,给出了不同参数的拓扑关系及推演算法,并探讨了其载荷谱的推演规律。基于离散正则化方法(DRM),采用Morozov偏差原则与Newton迭代格式相结合的手段选取正则参数,重构截齿载荷谱,并提取载荷谱特征参量。结果表明:重构与推演的截割载荷谱特征易于辨识和提取,截割能量主要处在1～3Hz的低频带上;楔入角β=45°与β=40°的载荷均值关系为F45°≈1.4F40°,其相关系数R=0.8603;重构与推演的算法对截齿截割载荷的反演定量求解十分有效。     

三维接触问题的非光滑算法

给出了一种非光滑算法直接用于求解三维摩擦接触问题的不可微非线性互补模型,不再对模型进行光滑化处理,使算法更加简单。文中对非光滑算法的收敛性给出了严格的数学证明,数值实验表明该算法列式简单,但与光滑化算法同样有效。     

机载武器极区传递对准算法

针对高纬度地区经线收敛导致以真北向作为航向参考的导航算法失效的问题,并结合战机在极区内具备正常导航和作战能力的需求,提出了基于格网导航力学编排的极区传递对准算法.首先以格林威治子午线作为航向参考,在此基础上定义了格网导航坐标系同时给出了格网导航力学编排及其误差方程,解决了极区内相对经线定向定位困难的问题.其次在格网导航坐标系下设计了主子惯导信息的“速度+姿态”匹配传递对准算法,估计和校正弹载子惯导的误差.仿真结果表明,该算法可避免载机进行费时的方位机动,仅需一次简单的摇翼机动就可令弹载惯导5 s 内方位对准精度达到5',为战机在极区内仍具有导弹攻击能力提供了工程应用参考.     

短波近似的保辛算法

WKBJ短波近似是最常用的有效求解方法之一。保守体系的微分方程可用Hamilton体系的方法描述,其特点是保辛。保辛给出保守体系结构最重要的特性。但WKBJ短波近似却未曾考虑保辛的问题。WKBJ近似可用自变量坐标变换,然后再给出其保辛摄动。数值例题展示了本文变换保辛算法的有效性。     

Influence of elastic compliance of links on the dynamics and accuracy of a manipulating robot with rotational and translational joints

We pose problems of dynamic and kinematic control of spatial motions for multilink manipulating robot with elastic links and with rotational and translational joints. These problems are reduced to solving a system of ordinary and partial differential equations of hybrid type in the independent variables. We use a numerical integration algorithm for such systems which was developed earlier for manipulating robots with elastic links of anthropomorphic type. We discuss the difficulties arising in mathematical simulation of manipulating robots with simultaneously rotational and translational joints and with elastic links. To perform a comparative analysis and estimate the positional accuracy for the center of mass of the weight transported by the manipulator, we pose problems of dynamic and kinematic control of spatial motions of a manipulating robots with rigid links and with rotational and translational joints. The resolving equations obtained in this case are based on the Lagrange formalism of the second kind. By way of example, we present the solution of dynamic control problems for elastic and rigid two-link manipulators with one translational and two rotational joints.     

多机械臂搬运同一物体的协调动态载荷分配

针对多机械臂共同搬运同一物体形成闭运动链的协调系统,研究了多机械臂协调动态载荷分配存在冗余情况下的实时性分配方法文中提出了“应集中惯性质量棒”的概念,进而根据各机械臂的承载能力,采用线性加权的方法发展了一种操作物体质心处零内力的载荷分配原则。最后给出了系统载荷分配的解析表达式,有效地解决了多机械臂搬运同一物体时其动态载荷分配的实时性问题。     

Kinematics of single-loop mechanisms and serial robot arms: A systematic approach

This paper presents a systematic approach, based on displacement group properties, to the kinematic analysis of spatial linkages with one closed loop and to the solution of the inverse kinematic problem for robot manipulators. By using the proposed approach, a set of kinematic chains can be determined such that a first closure equation with one unknown can be derived directly and explicitly. Then the remaining closure equations are obtained: it is proved that they can be expressed in triangular form. The basic algorithms used to solve these equations in closed form are also presented. For each algorithm, the conditions of applicability, the initial information required, the results, the type and form of equations, and the maximum number of solutions are given. The proposed approach is well suited to the symbolic explicit solution of the inverse kinematic problem of a wide range of robut mechanisms. An example of its application is given.

---

Sommario Il lavoro presenta un appreceio sistematico, basato sulle proprietà dei gruppi di spostamento, all'analisi cinematica di posizione di meccanismi spaziali ad una maglia e alla cinematica inversa di robot manipolatori seriali. L'approccio consente di determinare un insieme di catene cinematiche per le quali può essere scritta e risolta direttamente una prima equazione di chiusura in una sola incognita. Viene successivamente dimostrato esaustivamente che, per tali catene, le successive equazioni di chiusura possono essere espresse e risolte in forma triangolare. Inoltre sono presentati gli algoritmi di base utilizzabili per la soluzione del problema posto. Per ciascuno di essi sono dati: le condizioni di applicabilità, l'informazione iniziale richiesta, i risultati ottenuti, il tipo e la forma delle equazioni e il massimo numero di soluzioni possibili. L'approccio presentato è utilizzabile per la soluzione simbolica esplicita, manuale o automatica, di un esteso insieme di meccanismi per robot. Viene dato un esempio di uso del metodo.

     

Mobility and spatiality of parallel robots revisited via theory of linear transformations

Parallel robots are extensively used for various applications including manipulation, machining, guiding, testing and control. The mechanical architecture of parallel robots is based on parallel mechanisms in which a mobile platform is connected to a reference element by at least two legs. Mobility and spatiality are the main structural and kinematic parameters of a parallel robot. These two parameters are defined via the theory of linear transformation and can be easily determined by inspection using the definitions, properties and theorems introduced in this paper. An analytical method to compute these parameters is also presented just for verification and for a better understanding of their meanings. The new formalism presented in this paper is based on spatiality of an elementary open kinematic chain and relative spatiality between two elements of a closed kinematic chain. As far as we are aware, this paper demonstrates for the first time a new formula for calculation of general (full-cycle) mobility of parallel robots that overcomes the drawbacks of Chebychev–Grübler–Kutzbach's mobility criterion largely used for mobility calculation of multi-loop mechanisms. This new formula is easily applicable to parallel robotic manipulators with elementary or complex legs and mobility calculation does not involve the setting up of instantaneous constraint systems associated to the parallel mechanism.     

Nonlinear robust and optimal control of robot manipulators

In this paper, we propose a new optimal control method for robust control of nonlinear robot manipulators. Many industrial robot systems are required to perform relatively large angular movement with sufficient accuracy. In real circumstances, highly nonlinear manipulator dynamics and uncertainties such as unknown load placed on the manipulator, external disturbance, and joint friction make the precise control of manipulators a very challenging task. The main contribution of this work is to develop a new robust control strategy to accomplish the precise control of robot manipulators under load uncertainty using a nonlinear optimal control formulation and solution. This methodology is based on the underlying relation between the robust stability and performance optimality. A class of robust control problems can be transformed to an equivalent optimal control problem by incorporating the uncertainty bounds into the cost functional. The θ-D optimal control approach is utilized to find an approximate closed-form feedback solution to the resultant nonlinear optimal control problem via a perturbation process. Numerical simulations show that the proposed robust controller is able to control the robot manipulator precisely under large load variations.     

一类关节型机器人操作手的迭代学习控制

针对一类机器人操作手，提出了一种设计机器人操作手迭代学习控制的新方法，与其它现有的迭代学习控制方法相比较，当折合到机器人操作手关节轴上的驱动装置惯量足够大时，可以用驱动装置的动力学特性参数确定学习增益，克服了传统迭代学习控制方法由多次尝试选取学习增益的缺点，增强了实用性。在实际机器人操作手上的实验结果充分表明了新提出方法可获得较高学习速度和控制精度的优越性。     

The Kinematics and the Full Minimal Dynamic Model of a 6-DOF Parallel Robot Manipulator

In this paper we present a particular architecture of parallel robots which has six-degrees-of-freedom (6-DOF) with only three limbs. The particular properties of the geometric and kinematic models with respect to that of a classical parallel robot are presented. We show that inverse problems have an analytical solution. However, to solve the direct problems, an efficient numerical procedure which needs to inverse only a 3 × 3 passive Jacobian matrix is proposed. In a second step, dynamic equations are derived using the Lagrangian formalism where the joint variables are passive and active joint coordinates. Based on the geometrical properties of the robot, the equations of motion are derived in terms of only nine coordinates related by three kinematic constraints instead of 18 joint coordinates. The computational cost of the dynamic model obtained is reduced by using a minimum set of base inertial parameters.     

Free-floating closed-chain planar robots: Kinematics and path planning

The study of free-floating manipulators is important for the success of robotics program in space and in the design of innovative robot systems which can operate over a large workspace. In order to study the fundamental theoretical and experimental issues encountered in space robotics, a closed-chain planar manipulator was built at Ohio University (OU) which floats on a flat table using air bearings. Due to the absence of external forces in the plane of the table and couples normal to this plane, the linear momentum in the plane and the angular momentum normal to this plane are conserved. It is well known that the linear momentum equations are holonomic while the angular momentum equation is nonholonomic. Due to this nonholonomic behavior, the path-planning schemes commonly used for fixed-base manipulators do not directly apply to free-floating manipulators. In this paper, we present an algorithm for motion planning of planar free-floating manipulators based on the inverse position kinematics of the mechanism. It is demonstrated that the inverse position kinematics algorithms, commonly used for fixed-base manipulators, can be successfully applied to free-floating manipulators using an iterative search procedure to satisfy the nonholonomic angular momentum constraints. This procedure results in paths identical to those predicted by inverse rate kinematics. The inverse position kinematics algorithm is then used to avoid singularities during motion to result in successful paths. The results of the simulation of this algorithm using parameter estimates of the OU free-floating robot are presented.     

空间机器人双臂捕获卫星力学分析及镇定控制

随着航天技术的发展,空间机器人要求具有对非合作卫星的在轨捕获能力. 双臂空间机器人与单臂空间机器人相比在这方面显然更具有优势. 然而由于太空环境的复杂性,使得空间机器人双臂捕获非合作卫星操作过程的动力学与控制问题表现出下述特点:非完整动力学约束,动量、动量矩与能量传递变化,捕获前后结构开、闭环变拓扑,与闭环接触几何、运动学约束多者共存. 因此空间机器人双臂捕获卫星技术相关动力学与控制问题变得极其复杂. 为此,讨论了双臂空间机器人捕获自旋卫星过程的动力学演化模拟,以及捕获操作后其不稳定闭链混合体系统的镇定控制问题. 首先,利用拉格朗日第二类方程建立了捕获操作前双臂空间机器人的开环系统动力学模型,利用牛顿-欧拉法建立了目标卫星的系统动力学模型;在此基础上基于动量守恒定律、力的传递规律,经过积分与简化处理分析、求解了双臂空间机器人捕获目标卫星后受到的碰撞冲击效应,给出了合适的捕获操作策略. 根据闭链系统的闭环约束几何及运动学关系获得了闭合链约束方程,推导了捕获操作后闭链混合体系统的动力学方程. 最后基于该动力学方程针对捕获操作结束后失稳的闭链混合体系统,设计了镇定运动模糊H_∞ 控制方案. 提出的方案利用模糊逻辑环节克服参数不确定影响,由H_∞ 鲁棒控制项消除逼近误差来保证系统控制精度;通过最小权值范数法分配各臂关节力矩,以保证两臂协同操作. 李雅普诺夫稳定性理论证明了系统的全局稳定性. 最后通过数值仿真实验模拟、分析了碰撞冲击响应,并验证了上述镇定运动控制方案的有效性.      

链状柔性多体机器人系统动力学研究

本文基于Ｊｏｕｒｄａｉｎ变分原理建立了具有链状拓扑结构柔性多体机器人系统动力学通用模型，用在一致质量有限单元法及正则模态分析基础上引入的模态坐标描述构件的弹性形，用独立坐标描述相邻板件间的大位移运动，每个铰容许１－６个自由度，组强非线性惯性耦合的封闭形式的系统动力学微分方程组，文末对单弹性臂和双弹性臂机器人操作手进行动力学仿真。