基于ZUPT的新型月面导航处理方法

A novel ZUPT-based processing method for lunar surface navigation

关键词:  零速更新  多模型交互  惯性导航  月面导航

Key words:  Zero velocity update(ZUPT)  Multi-model interaction  Inertial navigation  Lunar surface navigation

摘要:

为了解决探测器在月面工作时惯性导航系统的误差累积问题，克服现有研究对环境敏感或是需要定期停止才能进行误差修复的短板，结合探测器的行驶行为，将探测器使用零速更新（ZUPT）原理修正误差的过程分解成了无ZUPT、完全ZUPT、水平动态ZUPT和垂向动态ZUPT 4种子模型的叠加，并分别推导了它们的系统矩阵和测量矩阵。针对月面导航系统提出了一种新型多模型零速更新（MMZUPT）算法。该算法应用多模型交互原理，允许多种更新模型同时工作，可以通过计算每个模型的权重在每个定位历元输出最优导航结果并约束误差。在精心挑选的受限环境中对月面导航进行了模拟实验，结果表明，所提MMZUPT方法无需频繁地主动停车以满足传统ZUPT使用边界就可以取得良好的效果，并且通过对包含错误的校准信号加以识别和利用可以获得更好的性能。

Abstract:

To solve the problem of error accumulation in the inertial navigation system when the rover is working on the lunar surface, and to overcome the shortcomings of the existing studies that are sensitive to the environment or the rover need to be stopped periodically for error fixing. Combining with the behavior of the rover, the error correction process of the rover using the principle of zero velocity update (ZUPT) is decomposed into four sub-models, namely, the no-ZUPT, the full-ZUPT, the horizontal dynamic ZUPT and the vertical dynamic ZUPT, and their system matrices and measurement matrices are separately derived. A novel multiple model zero velocity update (MMZUPT) algorithm is proposed for the lunar surface navigation system. Applying the principle of multi-model interaction, this novel navigation algorithm allows multiple zero velocity update models to work simultaneously, and can output the optimal navigation results and constrain the cumulative errors by calculating the weights of each model at each positioning epoch. The simulation experiments of lunar surface navigation in a carefully selected constrained environment show that the proposed MMZUPT approach achieves good results without frequent active stops to meet the traditional ZUPT usage boundaries, and that better performance can be achieved by identifying and exploiting calibration signals that contain errors.