论文部分内容阅读
本文研究交互式多模型滤波器(IMMF).IMMF对于动态模式具有随机突变的一类混合系统的估值具有良好的性能.本文首先给出IMMF的数学描述,揭示IMMF的机理特征,针对IMMF的结构特点,以PD-100多机处理系统为基础,研究了IMMF并行实现的处理器拓朴结构、IMMF任务划分、分配和IMMF并行映射实现,给出了IMMF并行实现算法.在PD-100系统上的仿真表明,本文的并行算法具有加速比线性好,并行饱和度低和并行效率较高的特点,是IMMF实时实现的一种理想的并行处理方案.
This article studies interactive multimode filters (IMMFs). IMMF has good performance for the estimation of a class of hybrid system with random mutation in dynamic mode. In this paper, firstly, the mathematical description of IMMF is given to reveal the mechanism features of IMMF. According to the structural characteristics of IMMF, based on PD-100 multiprocessing system, the processor topological structure, IMMF task allocation, IMMF parallel mapping implementation, given IMMF parallel implementation algorithm. The simulation on the PD-100 system shows that the parallel algorithm in this paper has the characteristics of good acceleration ratio, low parallel saturation and high parallel efficiency. It is an ideal parallel processing solution for real-time IMMF implementation.