中文摘要：

针对期望模式修正（EMA）变结构多模型估计中基础模型集固定不变而扩张模型修正能力有限等问题，提出了一种改进的EMA变结构多模型算法（M-EMA）．该算法将所用模型集合构造为混合网格结构，并引入自适应网格（AG）技术和可能模型集（LMS）技术，用于混合网格中修正模型网格与基础模型网格的生成，从而使参与状态估计的模型集合更加接近于系统真实模式，达到优化模型集合的目的．仿真实验证明，该算法有效提高跟踪精度和稳定性，同时减少了对目标机动方式与模型集合拓扑结构设计的依赖．

英文摘要：

A special kind of algorithm, known as an amendatory expected mode augmention algorithm （ M-EMA）, was proposed to avoid the fixed structure of a basic model set and the limited amendatory capability of an augmented model. The M-EMA constructed the model set of a hybrid model grid at the same time by introducing the tech-niques of the adaptive grid （AG） and the likely-model set （LMS） to produce the amendatory model grid and basic model grid. The M-EMA produced a model set which was used in current time resembling more closely the real sys- tem mode, and the model set was optimized. The result of the simulation indicates that the performance and stability of the M-EMA algorithm was higher than EMA and LMS, and the M-EMA algorithm noticeably weakened the in- fluence of the maneuvering style and the structure of the model set.