多模型自适应滤波及其应用研究

发布时间：2018-12-20 22:01

【摘要】：随着深空探测任务的增加,自主导航技术变得越来越重要。对于转移轨道上的探测器而言,由于它离太阳与各个行星的距离比较远,近地轨道上的自主导航方法无法满足其导航要求,此时天文导航是非常有效的方法。天文导航是一种全自主的方法,具有精度高、误差不随时间而累积、抗干扰能力强以及可提供位置姿态等信息的优点,非常适合于远距离飞行、外界环境复杂多变、飞行时间长的导航任务。在火星探测器的转移轨道上,地球与火星轨道之间有很多近地小行星,我们可以利用观测小行星获得信息来确定探测器的导航信息。在此,我们采用四体模型作为导航系统的状态模型。由于采用小行星图像信息和小行星星光矢量作为观测量,会出现姿态估计误差和敏感器指向误差;而采用星光角距作为观测量可以避免这两方面对导航精度的影响,故观测模型采用星光角距作为观测量。由于导航系统的状态方程和观测方程不可避免地出现误差,如果想获得更精确的状态估计值,就得用滤波估计方法系统的状态量进行估计。由于深空探测环境复杂多变,导航系统的状态模型噪声时刻变化,与单一模型相比,多模型自适应估计方法通过一组并行滤波估计器进行估计,并且时刻计算模型概率,符合这种过程噪声序列多变的需求,能达到自适应的效果。本文研究了多模型自适应估计与扩展卡尔曼滤波/无迹卡尔曼滤波相结合的方法,形成多模型自适应扩展卡尔曼滤波与多模型自适应无迹卡尔曼滤波这两种方法,并将其用于基于小行星观测信息的自主天文导航系统上。通过仿真实验,将基于单一模型的天文导航系统与基于多模型的天文导航系统进行了详细的比较,说明引入多模型自适应方法,可增强系统对环境的适应性,明显提高天文导航系统的精度、连续性及可靠性。
[Abstract]:With the increase of deep space exploration missions, autonomous navigation technology becomes more and more important. For the probe in the transfer orbit, because of its distance from the sun and the planets, the autonomous navigation method in the low Earth orbit can not meet its navigation requirements. At this time, astronomical navigation is a very effective method. Astronomical navigation is a fully autonomous method, which has the advantages of high precision, no accumulation of errors with time, strong anti-interference ability and the ability to provide position and attitude information. It is very suitable for long distance flight and complex and changeable external environment. A long flight navigation mission. There are many near-Earth asteroids between the Earth and Mars orbit in the transfer orbit of the Mars probe. We can use the observation asteroids to obtain information to determine the navigation information of the spacecraft. Here, we use the four-body model as the state model of the navigation system. Because of the asteroid image information and asteroid starlight vector as observations, there will be attitude estimation error and sensor pointing error. The influence of these two aspects on the navigation accuracy can be avoided by using the starlight angle distance as the observation quantity, so the starlight angle distance is used as the observation quantity in the observation model. Due to the inevitable errors in the state equation and observation equation of navigation system, if we want to obtain more accurate state estimation, we have to estimate the state quantity of the system by using the filter estimation method. Because of the complex and changeable deep space exploration environment, the noise time of the state model of the navigation system changes. Compared with the single model, the multi-model adaptive estimation method is estimated by a set of parallel filter estimators, and the probability of the model is calculated at the same time. In accordance with the requirements of the process noise sequence, the adaptive effect can be achieved. In this paper, the method of combining multi-model adaptive estimation with extended Kalman filter / unscented Kalman filter is studied to form two methods: multi-model adaptive extended Kalman filter and multi-model adaptive unscented Kalman filter. It is used in autonomous astronomical navigation system based on asteroid observation information. Through the simulation experiment, the astronomical navigation system based on single model and the astronomical navigation system based on multi-model are compared in detail. It is shown that the adaptive method of multi-model can enhance the adaptability of the system to the environment. The accuracy, continuity and reliability of the astronomical navigation system are obviously improved.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN713

【相似文献】