改进的模型集自适应变结构多模型算法

针对传统固定结构的交互式多模型算法模型集无法覆盖目标的所有运动模型,提出一种改进的模型集自适应变结构多模型算法。该算法在量测信息的基础上,采用Kullback-Leiber规则对模型集中各模型与目标运动模型匹配程度进行判别,删减匹配程度较低的模型;同时,采用改进的期望模型算法得到一个新的匹配程度高的模型集,并对原模型集中的模型进行更新;最后再与容积卡尔曼滤波器相结合,实现非线性目标跟踪系统的变结构多模型算法。仿真结果表明:该算法在保证系统计算量没有明显增加的基础上有效地增加了算法的自适应性能和系统的鲁棒性能,更好地满足了工程应用中对算法自适应性和实时性的要求。

Nonlinear Target Tracking Algorithm Based on Variable Structure Multiple-Model

Owing to the problem that the model set of the fixed interaction multiple-model algorithm could not include all the motion models during target tracking,an improved variable structure multiple-model algorithm based on model adaptation was proposed.Based on the measurement innovations,the Kullback-Leiber criterion was utilized to discriminate the matching between the subset and the real motion model.The worst subset then was terminated.By the reference from the Expected model algorithm,a new model was added to the new model set and the other subset models reserved then updated respectively.At last,a new filter called the Cubature Kalman Filter was used to estimate the states of nonlinear maneuvering target tracking system.The simulation results showed that the algorithm proposed in this paper enhanced the tracking system’s adaptation and robustness while the burden on computation was not increased obviously.That satisfied the requirement on algorithm adaptation and real time which was emphasized in project.