Research On Resource Management For Cognitive Radar And Radar Signal Processing Based On GPU

Cognitive radar is an important developing direction of the radar in the future,which changes the one-way pattern of traditional radar and develops a closed loop from transmitter to receiver,therefore cognitive radar is hopeful to significantly improve the combination property of radar.One of the important research contents about cognitiveradar is cognitive transmit,which is the significant difference from traditional radar.Resource management of cognitive transmit can make transmitter resource more effectively application on condition that system resource is not increasing,which plays an important part in increasing the comprehensive abilites of cognitive radar.Radar signal processing arithmertic is high computation cost,strong real-time and heavy computation,which has a high requirement for the platform of radar signal processing in computing power and transmission speed.In recent years,Graphical Processor Unit(GPU)has been successfully used in many fields because of its greater computing power and super parallelism,and combining GPU with high-speed and real-time radar signal processing is an important research direction.This paper studys the problem of resource management of cognitive radar transmitting terminal and radar signal processing based on GPU platform.The main contents are as follows:1.For the problem of the resource management with multi-targets,we propose a new algorithm of resource management with MIMO radar based on an improved bin-packing arithmertic.This algorithm takes the limited concurrent beam number of the real radar system into consideration.Compared with traditional bin-packing arithmertic,it increases the constraint of the targets number in the optimization process,so it not only optimizes the residual space of the bin,but also optimizes the number of the targets in the bin.The experiment results show that the improved arithmertic can significantly improve the efficiency of system transmit resource,and then improve the combination property of radar.2.Introduce the hardware of GPU and programming model of Compute Unified Device Architecture(CUDA).Firstly,we analyze the difference between GPU and Central Processing Unit(CPU)in hardware and performance.Then we introduce the programmable architectures of GPU,i.e,CUDA from programming model and storage model.Finally,the mixed programming between GPU and MATLAB is introduced.3.Study on the implement method of radar signal parallel processing on GPU platform,including pulse compression,clutter rejection,constant false alarm rate detection and particle filter.In the particle filter on GPU platform,aiming at the problem of the poor parallelism of resampling,the serial circulation condition of resampling is changed by SMSR,so the high parallelism of resampling can be achived and the speed of particle filter is markedly improved.The experiment results show that GPU can markedly enhance the speed of radar signal processing over CPU serial computing.The data is larger,the improvement of speed is more remarkable.