基于多核计算的雷达并行仿真结构

Parallel Structure Based on Multi-Core Computing for Radar System Simulation

针对顺序仿真结构下回波生成与信号处理环节软件仿真速度慢等瓶颈问题,提出一种基于多核处理器共享内存的多数据链路计算模型,通过构建多数据链路并行仿真的方法提升软件仿真效率。根据同一调度间隔内各雷达事件相互独立的特性,从数据划分、任务分配、时间同步及负载监测与度量等层面上进行阐述。仿真结果表明,该方法与传统的雷达串行仿真相比,数据帧处理平均时间可以降低37.5%,数据帧处理加速比曲线表现出良好的仿真加速特性,大大缩减雷达系统仿真时间。

To solve the bottle-neck problem of lower efficiency existed in radar echo generation and signal processing with serial simulation architecture, a multi-data links computing model based on multi-core memory-shared platform is proposed. This method could greatly promote simulation efficiency by taking advantage of multi-core. According to the independent characteristic between radar tasks in the same scheduling interval, the model takes data division, task allocation, time synchronization, and load monitoring with measurement into account to discuss its parallel characteristic. The Pentium（R） Dual-Core E5200 CPU with 2 GB memory is used to test the target scene with 20 batches. Simulation results demonstrate that, compared with serial simulation, the data frame average processing time of parallel model decreases 37.5% and the data frame processing speedup ratio curve has good acceleration performance. This parallel algorithm can reduce the simulation time greatly.