未知转换概率矩阵下的机动目标跟踪算法

机动目标跟踪过程中的转换概率矩阵往往是未知的,系统状态也将呈现非线性、非高斯、不完全观测的特点。传统的方法如交互多模型、广义伪贝叶斯算法等解决该类型问题的效果并不理想。将准贝叶斯法则和辅助粒子滤波算法相结合,提出了一种新的未知转换概率矩阵条件下的机动目标跟踪算法(QB-APF)。仿真结果表明,该算法与其他方法相比具有更高的滤波精度和较好的数值稳定性。

Study on maneuvering target tracking algorithm with unknown transition probabilities matrix

In practical situations,the transition probability matrix is always unknown during the process of maneuvering target tracking,and the system state is characterized by nonlinear,non-Gaussian and incompletely observed too.The traditional methods such as IMM and GPB deal badly with this kind of problem.Both the quasi-Bayesian algorithm and the auxiliary particle filter algorithm are combined to present a new maneuvering target tracking algorithm called QB-APF with unknown transition probability matrix.Simulation results demonstrate that the QB-APF algorithm improves filtering accuracy and has satisfied numerical stability compared with other algorithms.