基于卡尔曼滤波组合导航算法的计算量与精度分析

发布时间：2018-06-14 22:55

本文选题：惯性导航系统 + 精度　；参考：《哈尔滨工业大学》2014年硕士论文

【摘要】：最近几年,现代控制理论的不断发展完善以及计算机技术的进步,组合导航的应用也愈加广泛。组合导航从其得名可以知道就是让多种不同的导航系统通过算法的融合研究最终达到优势互补的目的。目前,国内外针对此项技术已经取得了不错的成效,先后推出了多种系列的组合导航系统,使其成为现如今最为重要的导航系统。本论文研究了惯性导航/全球定位组合系统中的关键技术,并重点研究后期不同数据融合算法对计算量与精度的影响因素。为了更好的得到长时间、高精度的参数,本论文中采用INS为基础系统同时GPS为辅助系统构成的INS/GPS组合导航,使得以上两种系统可以达到“互补”效果。进一步将惯性导航系统和全球定位系统得到的测量结果送入导航计算机中,进行数据滤波融合,最终得到导航参数。本论文中INS/GPS组合系统主要参数包含:物体姿态角,空间位置以及物体的空间速度。本论文着重讨论了INS/GPS组合系统的数据融合过程,并具体研究了四种不同的数据融合算法:基本卡尔曼滤波KF、扩展卡尔曼滤波EKF、无敏卡尔曼滤波UKF以及粒子滤波PF。首先从理论层次上针对四种不同的滤波算法做以下研究:分析同一种滤波系统下影响精度或计算量的因素并总结整体规律;分析同一种系统下不同滤波算法间计算量与精度关系,同时结合以上所有算法的研究结论对其应用条件和优缺点做出总结。最后使用matlab软件进行仿真验证,并给出仿真结果与分析。结果表明,在基本KF滤波中滤波步长,状态向量维数和初始值是现在仿真过程中影响精度的几大因素,通过仿真进行验证并得到最优的经验输入值。在同一种18维系统且滤波步长、初始值最优的情况下,分别运行组合导航的EKF和UKF滤波算法,结果发现UKF的平均使用时间是EKF时间的1.3倍,EKF的误差项是UKF的2倍。在经典粒子滤波中,主要是粒子数目影响精度和计算量,通过仿真进一步验证这一结论。总结以上四种滤波算法,总体表明计算量越复杂性的系统则非线性化越强,需要应用更为复杂的滤波融合算法,并且耗费时间更长,但精度也会成倍的提高。
[Abstract]:In recent years, with the development of modern control theory and the progress of computer technology, the application of integrated navigation has become more and more extensive. From the name of integrated navigation, it can be known that many different navigation systems can finally achieve the goal of complementary advantages through the fusion of algorithms. At present, many kinds of integrated navigation systems have been developed, which make them become the most important navigation system. In this paper, the key technologies of inertial navigation / global positioning integrated system are studied, and the influence factors of different data fusion algorithms on computation and precision are studied. In order to obtain long time and high precision parameters, the ins / R / GPS integrated navigation system based on ins and GPS is adopted in this paper, which makes the two systems "complementary". The results obtained from inertial navigation system (ins) and global positioning system (GPS) are fed into the navigation computer for data filtering and fusion, and finally the navigation parameters are obtained. In this paper, the main parameters of ins / GPS integrated system include: attitude angle, space position and space velocity of object. This paper mainly discusses the data fusion process of ins / GPS integrated system, and studies four different data fusion algorithms: basic Kalman filter KF, extended Kalman filter EKF, unsensitive Kalman filter UKF and particle filter PF. Firstly, four different filtering algorithms are studied from the theoretical level as follows: analyze the factors that affect the accuracy or computational complexity of the same filtering system and summarize the overall law; This paper analyzes the relationship between computational complexity and accuracy among different filtering algorithms in the same system, and summarizes the application conditions, advantages and disadvantages of all the above algorithms. Finally, the simulation results and analysis are given by using matlab software. The results show that the filter step size, the dimension of state vector and the initial value are the main factors that affect the accuracy of the basic KF filter. The results are verified by simulation and the optimal empirical input value is obtained. The EKF and UKF filtering algorithms of integrated navigation are run under the condition of the same 18 dimensional system with filtering step size and optimal initial value. The results show that the average time of use of UKF is 1.3 times that of EKF and the error term of EKF is twice that of UKF. In the classical particle filter, the number of particles mainly affects the precision and calculation, and the simulation further verifies this conclusion. Summing up the above four filtering algorithms, it can be concluded that the more complex the system is, the stronger the nonlinearity is, and the more complex filtering fusion algorithm is needed, the longer the time is, but the higher the precision is.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN967.2

【参考文献】