摘要
在矢量观测的基础上,针对单独的星敏感器定姿,提出了一种将粒子滤波(PF)和预测滤波相结合的姿态确定算法,通过设计粒子初始化,结合重要性采样、重采样和规则化等手段,成功地将姿态四元数作为状态粒子进行更新和传递,避免了状态方程的线性化和协方差矩阵的计算;利用预测滤波算法估计模型误差和姿态角速度,在保证滤波精度的同时,有效降低了粒子滤波器的维数。实验在某对地观测通用小卫星平台上进行,选取卫星自由飞行状态和飞轮控制对地稳定模式,分别对滤波器进行了仿真,实验结果验证了该算法对本质非线性、非高斯的卫星姿态估计问题具有快速的收敛性能和良好的稳定精度。该方法还为粒子滤波器的设计和无角速度敏感器测量的飞行器姿态确定提供了借鉴。
A particle filter fusing predictive filter based on vector observations is presented for satellite attitude determination using solely star sensor without gyro.By designing proper particles initialization,using sequential important sampling,resampling and regularization,the attitude quaternion can be directly updated and transferred as states,thus avoiding linearization of system state equations and covariance matrix calculation.By using predictive filter to estimate model error and attitude angular rate,the particles dimensions are reduced effectively without loss of estimation accuracy.A simulation study in a satellite platform of a lab,including earth-pointing mode and free flying mode,is used to verify the fast convergence rate and the steady accuracy of the algorithm,and results demonstrate that the algorithm is robust in nonlinear and non-Gaussian scenarios.The new method supplies experiences to particle filter design and other kinds of attitude determination without angular rate sensors as well.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2011年第5期1077-1085,共9页
Journal of Astronautics
基金
国家自然科学基金(40374054)
科技部863计划(2008AA12A216)
关键词
姿态估计
星敏感器
粒子滤波
预测滤波
矢量观测
Attitude estimation
Star sensor
Particle filter
Predictive filter
Vector observation
