Research On Real-time Performance Of Software Radar System

Traditional radar systems are usually based on embedded hardware platforms,and they also have some disadvantages when they are efficiently implemented.For example,they have low development efficiency,high maintenance cost,high upgrade cost,and difficult to expand functions.With the rapid development of digital technology and high-performance computing and the fact that traditional radars urgently need reforms,software radar has emerged.Software-based radars decouple traditional hardware and software of radars and transfer traditional radar information processing algorithms from embedded platforms to general server platforms.This has led to tremendous improvements in development,maintenance,expansion,and flexibility.Radar information processing algorithms are very sensitive to time requirements,so it is particularly important to verify the feasibility of radar information processing algorithms running on a common server platform.This article proceeds from the three aspects of the radar system algorithm module—pulse compression,moving target detection,constant false alarm rate,radar system operating environment and communication middleware.For the three algorithm modules of the radar system,based on its in-depth analysis,this article puts forward the following optimization directions: using C language to refactor the source code of three modules and aiming to improve the module performance in terms of loop,calculation,cache life,etc.;choosing Intel C/C++ Compiler which can optimize intermediate code and object code in depth to reduce execution time of executable programs;using FFTW3-the fastest Fourier transform library in the West to accelerate radar information processing algorithm;analyzing the time complexity of the algorithm in the module to optimize the algorithm time complexity for some algorithms;using Open MP to accelerate Radar information processing algorithms in parallel.For the operating environment of radar system,this paper mainly uses two methods to optimize it: the installation of kernel real-time preemption patch and kernel trimming.In order to improve the hard real-time capability of Cent OS 6.5,improve system interrupt response speed and reduce system delay,this paper applies Preempt-rt preemptive patch to system.In order to reduce the influence of irrelevant services and processes on the algorithm module in the system,this paper performs kernel trimming for Cent OS 6.5,so as to eliminate redundant and useless kernel modules.For communication middleware,this paper proposes to use DDS communication middleware to accelerate the software radar.By modularizing the radar information processing algorithm and spreading it over different nodes,the DDS communication middleware is used to realize data transmission between different nodes,and the radar information processing system for streamlined operation is implemented;due to the limitation of bandwidth and delay of traditional TCP/IP network model,the bandwidth of data exchange between different nodes of DDS is insufficient and the time delay is too large.This paper uses the Infiniband network card based on RDMA protocol to accelerate the network and finally meets the multi-node requirements for data transmission of Highbandwidth,low-latency.After deep optimization of these three levels,the time-consuming of the radar information processing algorithm is finally 0.796 ms,which can indicate that it is completely feasible to transfer the radar from the embedded platform to the general server platform.