航空学报 > 2016, Vol. 37 Issue (9): 2799-2808   doi: 10.7527/S1000-6893.2015.0349

空间非合作目标惯性参数的Adaline网络辨识方法

孙俊1,2, 张世杰1, 马也3, 楚中毅3   

  1. 1. 哈尔滨工业大学 航天学院, 哈尔滨 150001;
    2. 上海航天控制技术研究所 上海市空间智能控制技术重点实验室, 上海 201109;
    3. 北京航空航天大学 仪器科学与光电工程学院, 北京 100083
  • 收稿日期:2015-09-07 修回日期:2015-09-30 出版日期:2016-09-15 发布日期:2016-01-06
  • 通讯作者: 张世杰 男, 博士, 教授, 博士生导师。主要研究方向:小卫星技术。 Tel: 0451-86414117-8409 E-mail: sjzhang@hit.edu.cn E-mail:sjzhang@hit.edu.cn
  • 作者简介:孙俊 男, 博士研究生, 高级工程师。主要研究方向:航天器导航与控制。 Tel: 021-24183325 E-mail: sjlovedh@hotmail.com;马也 男, 硕士研究生。主要研究方向:空间机器人及航天器控制技术。 Tel: 010-82339013 E-mail: 2763251585@qq.com;楚中毅 男, 博士, 副教授。主要研究方向:空间机器人及航天器控制技术。 Tel: 010-82339013 E-mail: chuzy@buaa.edu.cn
  • 基金资助:

    国家自然科学基金(51375034,61327809);上海航天科技创新基金(SAST2015-075)

Adaline network-based identification method of inertial parameters for space uncooperative targets

SUN Jun1,2, ZHANG Shijie1, MA Ye3, CHU Zhongyi3   

  1. 1. School of Astronautics, Harbin Institute of Technology, Harbin 150001, China;
    2. Shanghai Key Laboratory of Space Intelligent Control Technology, Shanghai Institude of Spaceflight Control Technology, Shanghai 201109, China;
    3. School of Instrument Science and Opto-electronics Engineering, Beihang University, Beijing 100083, China
  • Received:2015-09-07 Revised:2015-09-30 Online:2016-09-15 Published:2016-01-06
  • Supported by:

    National Natural Science Foundation of China (51375034,61327809); Shanghai Aerospace Science and Technology Innovation Fundation (SAST2015-075)

摘要:

空间在轨操作中,航天器在对空间非合作目标的抓捕行动常常导致航天器本体的姿态和空间轨迹发生变化。为克服空间非合作目标对航天器本体动力学、运动学的影响,使控制系统做出精准及时的姿控策略调整,确保航天器正常在轨工作和轨迹姿态的高精度,需对抓捕的非合作目标的惯性参数进行辨识。针对传统辨识方法依赖广义逆求解导致的辨识过程运算量大,且数值容易产生剧烈振荡,造成辨识结果不稳定等不足,采用基于归一化最小均方(NLMS)准则的Adaline神经网络方法进行空间非合作目标惯性参数的辨识。首先,基于动量守恒理论建立抓捕后的航天器—机械臂—空间非合作目标系统模型;然后将辨识方程的系数矩阵作为网络的输入和输出,空间非合作目标的惯性参数作为神经网络的训练权重,基于迭代步长可变的NLMS准则实现对目标惯量参数的快速、准确辨识;最后,在构造的ADAMS/MATLAB联合仿真平台上进行了验证。仿真结果表明,基于NLMS准则的Adaline神经网络是一种快速、准确辨识目标惯量参数的有效方法。

关键词: 航天器, 非合作目标, 惯性参数, 神经网络, 辨识

Abstract:

During the operation in space, the spacecraft's attitude and trajectory are often affected by capturing the uncooperative target. In order to overcome the influence of uncooperative target on the dynamics and kinematics of spacecraft and ensure the high-precision attitude control strategy to be made and normal in-orbit condition, a process of identifying the inertial parameters of uncooperative targets should be accommodated. In order to avoid a large amount of computation induced by generalized inverse operation of traditional method in the identification process, which also causes severe vibration and unstability to numerical results, an Adaline neural network identification method based on normalized least mean square(NLMS) criterion is adopted. First of all, a system model composed of spacecraft, manipulator and uncooperative target is established based on the theory of momentum conservation. Then the weight parameters of the neural network representing the inertial parameters of uncooperative target are trained by the coefficient matrix of the identification equation as the input and output of the neural network via algorithm of NLMS with variable iterative step, and a fast and accurate process of identification is achieved. Finally, an ADAMS/MATLAB co-simulation platform is established, on which the proposed identification method is verified. The simulation results show that the Adaline neural network based on NLMS criterion is a fast and accurate method for identifying the target's inertia parameters.

Key words: spacecraft, uncooperative target, inertial parameter, neural network, identification

中图分类号: