面向弹载微波前视三维成像制导的异构可重构计算研究

As one of key technologies which could achieve precision guidance but need to be settled urgently, missile-borne radar forward-looking imaging has significant demand in military field. With the increasing development of new radar technology, further investigations on missile-borne forward-looking imaging algorithms have been conducted recently. However, such study is insufficient on how to realize real-time processing of forward-looking imaging with high computational complexity under the precondition of low load and low power consumption on resource-limited missile-borne platform. To break through this technical bottleneck, the project focuses on missile-borne heterogeneous reconfigurable computing methods. The research contents of this project include : 1) optimization of missile-borne forward-looking imaging method under complex motion condition and core sub-circuit IP design; 2) heterogeneous system-oriented software and hardware collaborative design and verification of missile-borne forward-looking imaging and accurate guidance; 3) dynamic reconfigurable strategy based on autonomous task planning and low power and fault tolerance design. This project will introduce the heterogeneous reconfigurable computing architecture into missile-borne signal processing platform and break through the basic problems as real-time optimization of missile-borne forward-looking microwave imaging algorithm, heterogeneous software/hardware task scheduling and efficient allocation of reconfigurable resource. It can provide theoretical basis and technical support for the improvement of forward-looking imaging and the development of high energy efficiency computing architecture, and thus has important research significance and practical value.

弹载雷达前视成像是实现导弹精确制导亟需解决的关键问题之一，在军事领域有着重大需求。随着新体制雷达的不断发展，弹载前视成像算法的研究逐渐深入。然而在资源受限的弹载平台，如何在确保设备小负荷、低功耗的前提下实现高计算复杂度的前视成像实时处理，这方面的研究仍非常欠缺。为了突破此瓶颈，本项目拟开展弹载异构可重构新型计算体系与方法研究。本项目的主要研究内容包括：1）复杂运动条件下弹载前视计算成像算法实时优化与核心子电路IP设计；2）面向前视成像制导的弹载异构体系软硬件协同设计与验证；3）基于自主任务规划的动态可重构策略和低功耗与容错设计。本课题将异构可重构计算体系引入弹载信号处理平台，重点突破弹载微波前视成像算法实时优化、异构体系软硬件任务调度以及可重构资源高效配置等基础问题，为促进雷达导引头前视成像方法完善以及高能效计算体系结构的发展提供理论依据和技术支撑，具有重要的研究意义和实用价值。

弹载雷达前视成像是实现导弹精确制导亟需解决的关键问题之一，在军事领域有着重大需求。然而在资源受限的弹载平台，如何在确保设备小负荷、低功耗的前提下实现高计算复杂度的前视成像实时处理是一个重大难题。本项目开展的主要研究内容包括：1）复杂运动条件下弹载前视计算成像算法实时优化与核心子电路IP设计；2）面向前视成像制导的弹载异构体系软硬件协同设计与验证；3）基于自主任务规划的动态可重构策略和低功耗与容错设计。取得研究成果包括：1）优化了空时随机辐射场前视稀疏成像算法，完成了核心电路参数化IP设计；2）提出了异构可重构设计思路，构建了弹载多任务自适应动态重构集成微系统；3）建立了异构可重构新体系软硬件统一编程模型，实现了侦察干扰探测成像多任务计算资源自适应匹配。4）构建了典型的前视成像精确制导试验场景，验证了成像制导算法的实时性和有效性。本项目融合了制导控制、算法优化、电路设计、计算架构、可重构资源优化配置和软硬件协同设计几大要素，突破了弹载前视成像算法映射以及弹载多任务规划与异构可重构资源动态调度规划等关键技术，形成了相关的软硬件成果。该研究具有重要的研究意义和实用价值，基于该成果，课题组相继获得了国家级重点项目“精确制导集成微系统设计新技术”和“弹载侦干探通一体化信号处理集成微系统”支持，相关理论和技术正逐步往航天科工集团二院、三院等单位转化，有力支撑了弹载精确制导信号处理和多功能集成微系统的技术发展。