Design And Implementation Of The Radar Signal Real-Time Processing System Based On Embedded Reinforcement GPGPU Simulation Platform

Implementing General-Purpose Computation on GPU (GPGPU) has become one of the focuses of science and technology research areas at home and abroad in recent years. Such as the United States, Italy and other countries abroad, the GPGPU technique has been maturely used in Unmanned Aerial Vehicle (UAV) system, multiple aperture radar system, and other areas of the military in a few years ago. In the current domestic military field, the number of practical engineering projects which perform general parallel accelerating via GPGPU is relatively limited. One of the basic reasons for this situation is the platform built of ordinary commercial graphics cannot meet the actual demands of the resistance to severe environment and the need of the embedded reinforcement of system, the other is that the design of the double or more GPU programming model needs higher professional abilities for designers in relevant areas, while the OpenCL specification has not been widely accepted by the engineers and technicians in the field of radar applications.The work of this thesis is based on GPU embedded reinforcement simulation platform developed by siasun company. On the basis of OpenCL programming method, this thesis offers a parallel acceleration solution for radar signal processing and provides a method of system optimization design. Aimed at solving the previous encountered bottleneck during performing radar signal processing operations, this thesis makes full use of the high efficiency and low power consumption of GPU in large-scale parallel computing and provides a set of mature solutions and operation models to shorten the research and development cycle in the field of GPU computing for engineers and technicians in related fields. The design of GPU parallel model for signal processing, load balancing and the optimization of parallel system is the key research contents of this thesis.Based on the model of radar signal processing system and according to the signal processing process, this thesis realizes the parallel accelerated computation for pulse compression module, moving target display module, constant false alarm circuit module and target extraction module respectively under OpenCL software framework.It also takes into account the task scheduling allocation and load balancing when using multiple GPUs at the same time in this thesis. After optimization of the whole system, the expected performance goal is to get the acceleration of 10 to 50 times than ever before. The author’s main work is as follows:(1) Analysis and design of the radar real-time signal processing;(2) Design and implementation of the heterogeneous parallel processing system based on GPU;(3) Task scheduling and load balancing for multiple GPUs;(4) Design and optimization methods for the key algorithms;(5) Optimization of the entire system based on the hardware architecture.The research of this thesis breaks through the bondage of previous serial programming modelin Radar signal processing field and achieves the efficient implementation of radar signal processing model under parallel heterogeneous platform.