一种新的针对模型误差所致滤波发散的补偿方法

摘要

Model errors is a main cause of Kalman Filtering divergence. Two compensation methods for filteringdivergence are analysed in this paper. A new algorithm named fading-limited memory is proposed throughintroducing a threshold to combined these two methods. A numerical example testify that the new algorithm has abetter performance in compensating filtering divergence caused by model errors, and it can choose measurementsadaptively.%系统模型误差是导致卡尔曼滤波发散的一个很重要的原因,分析了衰减记忆法和限定记忆法两种对模型误差所致滤波发散的补偿方法,通过引入阈值可以将这两种方法进行有机结合,给出了一种新的衰减限定记忆算法.算例分析验证了衰减限定记忆算法在抑制滤波发散方面具有比衰减记忆和限定记忆法更好的性能,并能自适应调整算法所需观测数据数目.