新的非线性滤波算法在GPS/DR组合导航中的应用

A new nonlinear filter for GPS/DR integrated navigation system

为解决传统粒子滤波算法中样本贫化的问题,提出一种新的粒子滤波算法.在重要性采样过程中,利用最新测量值,结合UKF滤波来产生粒子滤波中的建议分布;同时在再采样过程中,用高斯混合模型表示后验状态密度,引入最大期望(Expectation Maximization,EM)算法来获得该后验状态密度的参数,从新的参数分布中进行采样得到样本粒子,取代传统的再采样过程.把新算法应用到车辆组合导航系统中,仿真结果表明新算法的有效性.

In order to solve the ＂ sample impoverishment＂ problem in traditional particle filter, a new method is proposed. An effective proposal distribution is obtained from unscented Kalman filter using the latest measurement; and a new resampling process is accomplished by sampling from a no- vel distribution based on Gaussian mixture model resulting from expectation-maximization （EM） algorithm. The effects caused by sample impoverishment are lessened. Test on the global position system/ dead reckoning （GPS/DR） integrated navigation system and simulation results demonstrate the effectiveness of this new method.