机动多目标跟踪并行算法研究

A Parallel Algorithm for Tracking Multiple Maneuvering Targets

分析了直角坐标系下交互式多模型概率数据关联（ＩＭＭＰＤＡ）算法的并行机理。将实现该算法的任务分配到高速并行仿真计算机上，得到了并行ＩＭＭＰＤＡ算法。研制完成了机动多目标跟踪并行仿真软件包，对并行算法性能进行了测试。仿真结果表明跟踪精度和并行效率良好。

Most of the classic algorithms in target tracking are serial algorithm. They cannot meet with the requirements of real time tracking of multiple maneuvering targets. We present a method to change a well known IMMPDA (Interacting Multiple Model Probabilistic Data Association) algorithm into a parallel one and implement it in a specific parallel computer. Fig.2 shows schematically the hardware structure of the parallel computer which has five node computers. In subsection 3.1, we studied the parallel mechanism of IMMPDA algorithms based on the computer structure in Descartes coordinates. In subsection 3.2, we gave the algorithm for the parallel IMMPDA. We also developed a software to implement the parallel IMMPDA (section 4). Table 1 gives the acceleration ratios and the efficiency of the parallel IMMPDA. Table 2 gives the run time of the parallel IMMPDA(1.770 s) and those of IMMPDA on Personal Computers of different configurations. Results show that our parallel IMMPDA is efficient and easy to implement.