在目标跟踪中,针对无迹卡尔曼滤波在高维状态下容易出现滤波精度下降甚至发散的问题,提出了一种自适应交互式多模型容积卡尔曼滤波算法。首先,将容积卡尔曼滤波引入到交互式多模型算法中,提高了算法在高维非线性情况下的滤波精度。然后,结合马尔科夫参数自适应思想,在模型概率更新阶段,利用后验信息修正马尔科夫概率转移矩阵,增大匹配模型的转移概率,进一步提高模型之间的切换速度。最后,在目标跟踪仿真中利用“当前”统计模型对算法进行验证,实验结果证明了算法的有效性。

To solve the problem that the filtering accuracy of unscented Kalman filter is tend to decrease or diverge in maneuvering target tracking, an adaptive interacting multiple model cubature Kalman filter(AIMMCKF) is proposed. Firstly, the cubature Kalman filter is combined with interacting multiple model to improve the filtering precision under high-dimensional and non-linear situation. And then, the adaptive Markov parameter method is utilized in model probability update step. The Markov transition matrix is revised by posterior information, thus the matching model probability is magnified and the switching time is decreased. Finally, two current statistical models are used in target tracking simulation to examine the new algorithm, simulation results demonstrate the availability of AIMMCKF.

国家自然科学基金资助项目