论文部分内容阅读
为了解决地形辅助导航中面临的高维、强非线性问题,提出了基于Rao-Blackwell框架的RB-GSPF算法。该算法将原系统中的线性Gauss子结构分离出来,使用经典的Kalman滤波器处理,而剩下的强非线性部分通过Gauss和粒子滤波器处理,这种结构上的分解既发挥了Kalman滤波器对于线性Gauss系统的最优性,又利用了GSPF算法结构上的优点。理论及实验分析表明:该算法与粒子滤波器相比,在降维的同时提高了定位精度,减少了粒子数目;与Rao-Blackwellised粒子滤波器(RBPF)相比,其算法结构具有更好的并行性,从而在运算量上具有优势。
In order to solve the high dimensional and strong nonlinear problems in terrain-assisted navigation, an RB-GSPF algorithm based on Rao-Blackwell framework is proposed. The algorithm separates the linear Gauss substructure in the original system and uses the classical Kalman filter, while the remaining strong non-linear part is processed by Gauss and particle filter. The structural decomposition can not only play the roles of Kalman filter For the optimality of the linear Gauss system, the structural advantages of the GSPF algorithm are also taken advantage of. Theoretical and experimental analysis shows that compared with particle filter, this algorithm improves the positioning accuracy and reduces the number of particles while reducing the dimension. Compared with RBOF, the proposed algorithm has better algorithm structure Parallelism, which has an advantage in the amount of computation.