资源限制型可重构并行信息处理方法  

Resource-constrained parallel information processing methods based on reconfigurable technology

作　　者：陆智俊[1,2] 贲德[1] 毛博年[3] 

机构地区：[1]雷达成像与微波光子技术教育部重点实验室南京航空航天大学,江苏南京210016 [2]上海市空间智能控制技术重点实验室,上海201109 [3]中国科学院国家空间科学中心,北京100190

出　　处：《红外与激光工程》2016年第11期353-358,共6页Infrared and Laser Engineering

基　　金：国家自然科学基金(61271327)

摘　　要：针对立方体钠卫星GNC信息处理系统高计算性能与低功率消耗相矛盾的问题,提出了一种资源限制型可重构并行信息处理方法。该方法采用紧耦合可重构并行信息处理架构,将GNC信息处理中需要多次迭代计算且不适合CPU处理的复杂软件算法,以动态部分重构硬件电路单元(DPR)的方式实现,采用基于互斥量的多核并行可重构资源调度算法,通过多核CPU并行管理与调度共享的DPR单元,完成软件算法的硬件加速与优化。实验结果表明,该方法实现了立方星GNC信息处理系统的高效实时快速处理,与传统信息处理方法相比,可节约50%左右的功耗,可应用于计算资源极为有限的星上信息处理领域,具有很好的工程应用前景。Abstract： A method of reconfigurable parallel information processing in GNC was proposed for conflicting issues between high calculation performance and low power consumption in information processing of cubesats. An architecture of tightly coupled reconfigurable parallel processing was adopted in the method and complex algorithms which require many iterations and are not suitable for CPU operations in GNC information processing were achieved by dynamically partial reconfigurable unit of the hardware circuit （DPR）. A multi-core parallel reconfigurable resource scheduling algorithm based on mutex was adopted in the method and optimized hardware acceleration of the software algorithms was completed through multi-core CPU parallel calls by DPR unit. The experimental results indicate that real- time information processing of GNC is effectively achieved in cubesats and reduces power consumption up to 50%. The method can be widely used in the field of imformation processing for resource- constrained satellites and possesses a great application prospect.

关 键 词：动态部分重构 并行处理 立方体钠卫星 GNC 

分 类 号：V446.1[航空宇航科学与技术—飞行器设计]