A model-free method for attitude estimation and inertial parameter identification of a noncooperative target

Determining the attitude and inertial parameters of a noncooperative target is essential in an on-orbit servicing mission. Various methods based on machine vision have been proposed, but most of them require the 3D model of the target. This paper proposes a model-free method through sequentially registering point clouds captured by a depth camera. Our main contributions are the avoidance of the ambiguity in registration, and the combination of the multiplicative extended Kalman filter and the pose graph optimization to reduce the effect of measurement noise and drift error. A hardware experiment was performed to generate the sequence of point clouds of a three-axis free-floating target and validate our method. The result shows that the proposed method outperforms existing methods and effectively identifies the inertial parameters, including the normalized principal moments of inertia and the orientation of principal axes.     

Pose measurement and motion estimation of non-cooperative satellite based on spatial circle feature

Malfunctioned satellites have seriously threatened orbital safety, and the capture of these satellites is of great significance. The pose measurement and the motion estimation of the tumbling satellite is the premise of capture. In this paper, the docking ring of the satellite is identified, which is equivalent to a spatial circle. Combined with the nozzle feature, the pose duality of the spatial circle can be eliminated. And the measurement accuracy is improved by minimizing the reprojection error of the docking ring and the nozzle. Due to the symmetry of the docking ring, the measured pose has only five degrees of freedom, losing the degree of freedom of rotation around the normal vector. In the motion estimation algorithm, the observability of the tumbling motion is firstly analyzed, then an error-state Kalman filter with inertia ratio constraints is designed. To improve the convergence speed and stability of the filter, a rough estimation algorithm of filter initial value based on linear term extraction and particle swarm optimization is proposed. The effectiveness of the pose measurement and motion estimation method is verified by simulations.     

Motion prediction of a non-cooperative space target

Capturing a non-cooperative space target is a tremendously challenging research topic. Effective acquisition of motion information of the space target is the premise to realize target capture. In this paper, motion prediction of a free-floating non-cooperative target in space is studied and a motion prediction algorithm is proposed. In order to predict the motion of the free-floating non-cooperative target, dynamic parameters of the target must be firstly identified (estimated), such as inertia, angular momentum and kinetic energy and so on; then the predicted motion of the target can be acquired by substituting these identified parameters into the Euler’s equations of the target. Accurate prediction needs precise identification. This paper presents an effective method to identify these dynamic parameters of a free-floating non-cooperative target. This method is based on two steps, (1) the rough estimation of the parameters is computed using the motion observation data to the target, and (2) the best estimation of the parameters is found by an optimization method. In the optimization problem, the objective function is based on the difference between the observed and the predicted motion, and the interior-point method (IPM) is chosen as the optimization algorithm, which starts at the rough estimate obtained in the first step and finds a global minimum to the objective function with the guidance of objective function’s gradient. So the speed of IPM searching for the global minimum is fast, and an accurate identification can be obtained in time. The numerical results show that the proposed motion prediction algorithm is able to predict the motion of the target.     

基于先验子图检测的失效航天器SLAM方法

基于激光雷达的航天器位姿估计技术是当前在轨服务研究热点。针对失效航天器位姿估计,将通用的图优化SLAM技术应用到空间非合作目标的研究中。为解决SLAM算法在动态场景中产生累积误差问题,利用失效航天器自身运动特点,提出一种基于先验子图检测改进的SLAM算法。在该算法中,通过激光雷达和惯性测量单元分别采集失效航天器及周围环境的点云数据、服务航天器的运动信息,构建出服务场景下航天器的相对位姿图;再采用先验子图检测方法建立不连续的位姿节点间的约束关系;最后用约束信息对位姿图进行优化。仿真结果表明,相较于通用的SLAM算法的位姿估计,该方法减小了累积误差,提高了相对位姿估计精度,可以为后期的导航、控制等在轨任务提供信息。     

基于单目视觉的非合作目标相对位姿测量

为实现对空间姿态翻滚航天器的在轨服务与维护以及对空间碎片的清理，需对其进行精确的相对位姿测量。针对相对位姿测量问题，提出了基于单目视觉与卡尔曼滤波的相对位姿测量方法。通过对特征点匹配算法进行调查，采用了具有尺度不变性与旋转不变性的尺度不变特征变换算法(SIFT)和加速稳健特征算法(SURF)的特征点提取方法，并对二者进行了对比，得到了二者分别适用的工况条件。通过对Kalman滤波算法进行研究，引入了相机偏置矩阵，设计了Kalman滤波器，解决了单目相机的距离模糊问题，估计得到了非合作目标的相对位姿、主惯量比以及特征点位置信息。经过仿真，姿态角度估计误差在稳定后低于0.3°，相对位置估计误差在稳定后低于0.5m，相较于真值，误差小于1.67%，主惯量比估计误差在稳定后低于0.01，特征点位置误差在稳定后低于0.005m。在引入相机偏置条件后，滤波状态变量均收敛，并得到具有足够精度的估计，成功解决了单目相机深度信息缺失问题。     

基于截断最小二乘和半正定规划的空间非合作目标相对位姿估计

面向空间在轨服务、碎片清除重大应用需求,围绕空间光场复杂、空间目标尺度变化及相对姿态与位置强耦合等难题,提出一种基于截断最小二乘与半正定规划的空间非合作目标相对位姿估计方法。基于TOF相机,首先进行体素下采样,提取快速点特征直方图（Fast Point Feature Histograms, FPFH）特征,考虑目标的局部几何结构尺寸,计算两组FPFH特征的相似度并预测特征之间的对应关系,采用截断最小二乘和半正定规划方法对姿态、位置进行解耦,实现空间非合作目标相对位姿估计。使用非合作目标的3D模型进行仿真验证,仿真结果表明：该方法在三种不同高斯噪声强度下具有较好的估计特性,为后续的工程应用及在轨任务实施提供理论与技术支撑。     

Motion-planning and pose-tracking based rendezvous and docking with a tumbling target

Rendezvous and docking (RVD) with a tumbling target is challenging. In this paper, a novel control scheme based on motion planning and pose (position and attitude) tracking is proposed to solve the pose control of a chaser docking with a tumbling target in the phase of close range rendezvous. Firstly, the current desired motion of the chaser is planned according to the motion of the target. In planning the desired motion, the “approach path constraint” is considered to avoid collisions between the chaser and the target, and the “field-of-view constraint” is considered to make sure the vision sensors on the chaser to obtain tight relative pose knowledge of the target with respect to the chaser. Then, the difference between the chaser’s motion and the desired motion is gradually reduced by a pose tracking controller. This controller is based on the non-singular terminal sliding mode (NTSM) method to make the tracking error converge to zero in finite time. Since the chaser nearly moves along the desired motion and the motion is reasonable, (1) it could safely arrive at the docking port of the target with a suitable relative attitude, (2) it will be always suitably oriented to observe the target well, and (3) the magnitude of the needed control inputs are less than that in existing literatures. The numerical results demonstrate the above three advantages of the proposed method.     

基于视频的三维人体姿态估计

已有的三维人体姿态估计方法侧重于通过单帧图像来估计人体的三维姿态,忽略了视频中前后帧之间的相关性,因此,通过挖掘视频在时间维度上的信息可以进一步提高三维人体姿态估计的准确率。基于此,设计了一种可以充分提取视频时序信息的卷积神经网络结构,在获得高精度的同时也具有消耗计算资源小的优点,仅仅使用二维关节点的坐标为输入即可恢复完整的三维人体姿态。然后提出了一种新的损失函数利用相邻帧间人体姿态的连续性,来改进视频序列中三维姿态估计的平滑性,同时也解决了因缺少帧间信息而导致准确率下降的问题。通过在公开数据集Human3.6M上进行测试,实验结果表明本文方法相比目前的基准三维姿态估计算法的平均测试误差降低了1.2 mm,对于视频序列的三维人体姿态估计有着较高的准确率。      

基于特征点识别的头部姿态计算

为了提升使用随机回归森林进行头部姿态分析的精度,提出了一种基于特征点识别分析头部姿态的计算框架.考虑到高误差投票的干扰,该计算框架以随机森林的特征点识别为基础从而避免异常投票干扰,将头部姿态计算问题转换为空间鼻尖特征点和朝向特征点的识别问题.在随机森林的训练中,决策函数使用了高斯曲率和平均曲率作为图形特征,根据微分熵的信息增益在随机生成的决策函数库中搜索最优化决策函数.在训练完成的随机回归森林的叶子节点中,通过分析保存的样本数据,可以得到目标特征点的高斯分布估计.根据实验测试结果,在适当的阈值设定的情况下,该方法可以实现较高的识别成功率,使用曲率后明显提高了识别精度,能够在一定程度上处理有遮挡的数据,并且该方法已经成功应用于虚拟座舱的实时头部姿态分析计算系统.     

再谈一类挠性结构的约束及非约束模态解法

对于均质挠性梁附着于中心刚体上，且挠性梁具有尖端质量和尖端转动惯量的结构，该文分析了其运动方程的约束模态和非约束模态解，分别得到了相应的振型函数和普遍的频率方程；在此基础上定量分析了约束模态和非约束模态的频率随不同挠性梁／中心刚体惯量比的不同近似程度，为动力学建模和控制器设计提供了理论依据。     

转动惯量未知的非合作目标角速度估计方法研究

针对转动惯量未知的非合作目标的姿态角速度估计问题,将解算得到的非合作目标的惯性姿态作为测量信息,估计姿态角速度和姿态动力学参数（即转动惯量比）．首先,应用非线性控制系统的几何理论对待估计的状态扩展系统进行能观性分析．然后,利用UKF设计相应的滤波估计算法．仿真结果表明,本文所设计的方法能够精确估计出非合作目标的姿态角速度与转动惯量比．     

基于扩展卡尔曼滤波的舰机相对位姿估测

通过将基于扩展卡尔曼滤波的长序列图像分析方法与单目视觉技术相结合,把无人机自主着舰视觉导引中舰机间相对位姿的估测,转化为机载摄像机对着舰靶标平面3D位姿的实时估测问题.首先根据透视投影理论,建立了以摄像机的透镜中心为原点且Z轴与光轴重合的摄像机坐标系和世界坐标系,然后利用机载摄像机连续拍摄的靶标图像序列,选择描述相对运动的3个欧拉角、平移向量及它们的速度作为状态变量;由靶标角点的提取和帧间匹配,建立了反映着舰靶标上特征点的图像坐标和状态变量之间关系的观测方程,带入扩展卡尔曼滤波器,估测出舰机的相对运动参数.计算机数据仿真和基于DSP平台的半实物仿真试验验证了算法的有效性和鲁棒性.      

Prospects of using a permanent magnetic end effector to despin and detumble an uncooperative target

Space debris, such as defunct satellites and upper stages of rockets, becomes an uncooperative target after losing its attitude control and communication ability. In addition, tumbling motion can occur due to environmental perturbations and residual angular momentum prior to the object’s end-of-mission. To minimize the collision risk during docking and capturing of the tumbling target, a non-contact method based on the eddy current effect is put forward to transmit the control torque to the tumbling target. The main idea is to induce a controllable torque on the conducting surface of the tumbling target using a rotational magnetic field generated by a Halbach rotor. The radial and axial Halbach rotors are used to damp the spinning and nutation motions of the target, respectively. The normal and tangential force are evaluated concerning the relative pose between the chaser and the target. A simplified dynamic model of the nutation damping and despinning processes is developed and the influences of the asymmetrical principal moments of inertia and transverse angular velocity are discussed. The numerical simulation results show that the designed Halbach rotor stabilized the target attitude within an acceptable time. The electromagnetic nutation damping and despinning method provides new solutions for the development of on-orbit capture technology.     

一种结合深度学习的运动去模糊视觉SLAM方法

针对高动态环境下视觉同步定位与地图构建（Simultaneous Localization And Mapping,SLAM）系统的可靠性受运动模糊的限制,研究了一种基于生成对抗网络(Generative Adversarial Network,GAN)和AKAZE特征点的运动去模糊视觉SLAM方案。首先对因相机快速运动而产生的模糊图像进行AKAZE特征点的提取与检测,并根据特征点分布的丰富程度计算图像块权重,结合灰度图像的方差信息建立特征点与模糊程度之间的量化关系表;之后将达到模糊分数阈值的图像同步输入至改进GAN网络模型,该网络以端对端的形式恢复中心模糊帧的纹理信息,最后将输出的清晰图像重新进行位姿估计参与ORB-SLAM2后端优化过程。在公开数据集TUM上进行测试,对于受模糊影响较严重的数据集,方案可以明显降低相机轨迹估计的整体误差,同时维持较好的鲁棒性。     

基于三维点云重建的助推器位姿估计

助推器分离是运载火箭发射过程中的关键动作之一,常用的激光雷达姿态测量技术在助推器分离阶段受外界干扰严重,难以准确获得位姿。基于视觉的助推器位姿变化测量技术具有优秀的抗干扰能力,通过搭建三维点云重建网络,以图像为输入,三维点云为输出,在构建的助推器分离过程的图像 点云数据上进行了训练和测试,对测试重建的助推器点云使用主成分分析的方法完成了位姿的估算。测试结果表明,所建立的三维点云重建网络可以根据仿真图像数据,精确测量助推器分离阶段的位姿变化,在R2score指标下,对三维坐标的预测分数均在0.98以上,姿态角平均误差约为21°,预测分数则均在0.80以上。     

Wobble of a spin stabilized satellite with cross products of inertia and magnetic attitude control

Asymmetrical spin stabilized satellite dynamics in the vicinity of the required motion is considered. The principal axis of the maximum moment of inertia slightly deviates from its assumed direction in the satellite reference frame. This is formalized in the cross products of inertia. This inertial uncertainty results in a wobble, that is undesired angular velocity components perpendicular to the rotation axis, and oscillations of this axis near the required direction. The torque-free motion is investigated first. Expressions that explicitly relate satellite inertia parameters to wobble are provided. Wobble evolution under the action of magnetic damping control is analyzed next. Its amplitude approximate exponential decay behavior and residual unavoidable wobble level are derived. These expressions are compared with numerical simulation results of nonlinear equations of motion including various disturbance sources.     

水平起降两级入轨系统的质量估算方法及应用

针对近地轨道运输任务,考虑不同飞行阶段的推阻特性差异和不同发动机模态的比冲变化,提出了适用于水平起降两级入轨（TSTO,Two Stage To Orbit）天地往返运输系统的质量估算方法.以8 t有效载荷为任务需求,对比研究了涡轮基组合循环动力-可重复使用火箭（TBCC-RR,Turbine Based Combined Cycle-Reusable Rocket）、火箭基组合循环动力-可重复使用火箭（RBCC-RR,Rocket Based Combined Cycle-Reusable Rocket）和可重复使用火箭-可重复使用火箭（RR-RR,Reusable Rocket-Reusable Rocket）方案,分析了级间分离点、一级飞行器推阻比和一级飞行器结构质量分数等参数对设计结果的影响.研究结果表明,级间分离点设计对TSTO总体方案影响很大,若使用RBCC型飞行器作为第1级,建议在超燃冲压模态后即进行两级分离;TBCC-RR方案比RBCC-RR方案起飞总质量更小,但RBCC-RR方案一级飞行器结构质量更小;减小TSTO系统起飞总质量的最有效途径是减小飞行器的结构质量分数,其次是提高飞行器的推阻比.