1.国防科技大学 电子科学与工程学院，湖南 长沙 410073，sfwu1104@126.com
唐麒
1.国防科技大学 电子科学与工程学院，湖南 长沙 410073
魏急波
1.国防科技大学 电子科学与工程学院，湖南 长沙 410073;
2.中国电子科技集团公司第五十四研究所 通信网信息传输与分发技术重点实验室，河北 石家庄 050081
李卫
1.国防科技大学 电子科学与工程学院，湖南 长沙 410073;
3.中国人民解放军75753部队，广东 广州 510600
针对当前具有吞吐量和时延约束的软件无线电波形部署决策算法在内存开销方面考虑的不足，提出一个更为通用的波形部署决策系统模型，将波形的处理开销、内存开销和通信带宽开销都纳入到波形部署决策过程中，并设计了一种基于动态规划的波形部署决策算法以最小化波形对平台各种计算资源的占用总和。仿真结果表明，内存对波形部署决策具有重要影响，与未考虑内存开销的解决方案相比，平台支持的平均最大波形数目和平均处理资源利用率均下降40%左右。
新技术研究高校合作项目（KX142600035）；通信网信息传输与分发技术重点实验室资助项目（9140C170301140C17087）
1. College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China,sfwu1104@126.com
TANG Qi
1. College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China
WEI Jibo
1. College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China;
2. Science and Technology on Information Transmission and Dissemination in Communication Networks Laboratory, The 54th Research Institute of China Electronics Technology Group Corporation, Shijiazhuang 050081, China
LI Wei
1. College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China;
3. The PLA Unit 75753, Guangzhou 510600, China
Aimed at the problem that current deployment decision-making algorithms for software defined radio waveforms with throughput and latency constraints were paid little attention to the memory overhead, a more general system model for waveform deployment decision-making was proposed, by which the processing, memory and communication bandwidth were all incorporated in the process of waveform deployment decision-making. Besides, a dynamic programming-based algorithm for waveform deployment decision-making was devised to minimize the sum of various kinds of computing resource occupancies of waveform. Simulation results show that the memory has an important effect on the waveform deployment decision-making, while comparing with the algorithm that does not consider the memory overhead, the average maximum waveform number and the average processing resource usage supported by platform are both decreased about 40%.