基于修正PSO-UKF的SINS/GPS组合导航滤波算法  

SINS/GPS integrated navigation filtering algorithm based on modified PSO-UKF

作　　者：周卫东[1] 吉宇人[1] 乔相伟[1] 张鹤冰[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《中国惯性技术学报》2011年第5期565-570,共6页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金(60834005)

摘　　要：针对噪声时变特性引起滤波精度下降的问题,提出了一种基于修正粒子群技术(PSO)的自适应UKF算法。为了克服传统粒子群算法过早收敛,容易陷入局部最优的问题,基于粒子的适应值方差提出了一种惯性权值实时修正算法,有效改善了传统PSO算法。在使用新息序列对观测噪声进行实时跟踪的同时,通过构造合理的适应度函数将修正PSO算法和UKF滤波技术相结合,实现了对过程噪声统计特性变化的实时跟踪。针对SINS/GPS伪距组合进行了仿真实验。结果表明,该算法对时变噪声统计特性具有较强的自适应性,鲁棒性更强。在过程噪声及量测噪声发生变化的情况下,其对水平距离的估计精度比普通UKF算法提高了至少一倍,而水平速度的估计误差减小为不到原来的1/3。在对高度误差及天向速度误差进行估计时,普通UKF算法的估计误差很快发散,而PSO-UKF算法对高度的估计误差依旧能够保持在10 m以内。An adaptive UKF algorithm based on modified particle swarm optimization（MPSO） is proposed to solve the problem of filter divergence phenomenon caused by time-varying noise statistical property.To overcome the problem of premature convergence and local optimization in conventional PSO,an inertia weight dynamically adjusting algorithm based on particle fitness variance is derived to improve the PSO performance.By using the innovation sequence to track the observation noise in real time,an appropriate likelihood function is established,so that the MPSO can be combined with UKF to realize process noise statistical property tracking in real time.SINS/GPS pseudo-range integrated navigation system is simulated.The simulation results show that,the proposed algorithm has excellent adaptability and robustness for the statistical property of time-varying noises.In case there are changes in process noise and measurement noise,the estimation accuracy for horizontal distance is at least doubled,and the estimation error for horizontal velocity is less than one third of the general UKF algorithm.The estimations on height error and vertical velocity error show that,the conventional UKF algorithm diverges quickly,while the height estimation error of PSO-UKF can still be maintained to less than 10 m.

关 键 词：SINS/GPS 修正粒子群技术 无迹卡尔曼滤波 时变噪声 

分 类 号：U666.1[交通运输工程—船舶及航道工程]