基于CUDA的塔台模拟机冲突检测算法

塔台模拟机冲突检测算法是一种耗时大的并行算法。针对其导致塔台模拟系统核心服务器CPU负担过重的缺点,在常用冲突检测算法的基础上,提出一种基于统一设备构架（CUDA）的塔台模拟机冲突检测实现方案。首先介绍GPU并行运算的体系结构基础,并将基于卡尔曼滤波的目标物体跟踪技术的分层冲突检测算法移植到GPU。然后利用相同价格的CPU和GPU对比运算效果。实验结果表明：与相同算法的CPU实现方案相比,GPU实现方案将计算效率提高10～50倍。使用此方案,极大地减轻了核心服务器的负担,使塔台模拟机的性能得到质的提高。

Conflict Detection Algorithm of Tower Simulator Based on CUDA

The conflict detection algorithm of tower simulator is a large time-consuming parallel algorithm.For the defect which the algorithm causes CPU of tower simulation system＇s core server overload,on the base of common collision detection algorithms,provide a tower simulator collision detection implementations base on a unified device architecture（CUDA）.First,introduces the architectural base of GPU parallel computing,and transplant the conflict detection algorithm of tower simulator which based on Kalman filter object targeting and layered detection algorithm to GPU.Then compare the computing results between CPU and GPU which has the same price.Numerical experiments show that the implementation＇s computational efficiency on GPU increased by 10 to 50 times compared to the same algorithm implementations on CPU.Using this program,greatly reducing the burden on the core server,so that the performance of tower simulator quality improvement.