编队控制模型仿真是无人机编队设计中不可或缺的一部分。但当无人机规模较大时,仿真耗时问题突出。为解决仿真耗时问题,文中讨论了一种基于GPU的并行仿真方法并对所涉及的CUDA程序进行了优化。该方法设法将控制模型拆分成多个子模型,然后在多个仿真机上利用GPU同时求解子模型以提高效率。仿真结果验证了方法的正确性和高效性,同时也验证了GPU在大规模矩阵运算中的优势,且随着规模增大,优势愈加突出。

The simulation of formation control models has became an indispensable part of the design of multi-UAV formations. However, when the scale of UAV is large, the time-consuming problem of the simulation is outstanding. In response to this problem, this paper discussed a parallel simulation method based GPU and optimized the CUDA program. The method firstly splits the model into some sub-models, which will be solved on different simulators with GPU. The simulation results verify the correctness and efficiency of the method, and also verify the advantages of the GPU in large-scale matrix operations, and with the increase of scale, the advantages become more and more prominent.