作者：吕辰, 张晓明, 檀杰, 晁正正
作者单位：中北大学 仪器科学与动态测试教育部重点实验室, 山西 太原 030051
关键词：无人机;姿态解算;Cortex-M4F;DSP;浮点运算
摘要：
无人机进行飞行姿态测量解算的过程中，高采样率、较大解算量造成CPU负载过重，会影响姿态解算的实时性和准确性。针对此问题，提出一种基于Cortex-M4F内核的机载算法硬件加速技术。在分析无人机姿态解算算法特点的基础上，通过Cortex-M4F内核集成的DSP与浮点运算单元，对算法中三角函数、矩阵运算以及浮点数解算进行加速优化，并利用内核中CCM存储器代替SRAM，保证CPU数据存取时总线数据传输带宽的相对独立，实现无人机算法的高更新率与实时解算。实验结果表明，在实际机载飞行过程中进行算法的硬件加速，单周期解算时长减少至原时长的63.4%，数据更新率提高至500 Hz以上，在保证无人机姿态解算精度的同时，能够满足高动态数据采样的实时性要求。
Hardware acceleration technology for UAV attitude computation based on Cortex-M4F kernel
LÜ Chen, ZHANG Xiaoming, TAN Jie, CHAO Zhengzheng
Key Laboratory of Instrumentation Science & Dynamic Measurement, Ministry of Education, North University of China, Taiyuan 030051, China
Abstract: In the process of UAV flight attitude measurement and computation, the high sampling rate and large computation load will lead to the overload of CPU, which will affect the real-time and accuracy of attitude computation. Aiming at this problem, a kind of airborne algorithm hardware acceleration technology based on Cortex-M4F kernel is proposed in the paper. On the basis of analysis for UAV attitude computation algorithm characteristics, Cortex-M4F kernel integrated DSP and floating-point unit are adopted for acceleration and optimization of trigonometric functions, matrix operation and floating-point computation in the algorithm. Also, SRAM is replaced by CCM memorizer in kernel to ensure the relative independence of Bus data transmission bandwidth when accessing CPU data and realize the high update rate and real-time computation of the UAV attitude algorithm. The test results show that in the actual process of flight, the original computation period of a single cycle is decreased to 63.4% after hardware acceleration of the algorithm, and the data update rate is increased to more than 500 Hz, which can meet the real-time requirements of high dynamic data sampling while ensuring the accuracy of UAV attitude computation.
Keywords: UAV;attitude computation;Cortex-M4F;DSP;floating-point calculation
2018, 44(4): 106-111 收稿日期: 2017-09-20;收到修改稿日期: 2017-11-12
基金项目: 国家自然科学基金（51375463）
作者简介: 吕辰(1991-),男,山西太原市人,硕士研究生,专业方向为地磁导航。
参考文献
[1] MARINA H G D, PEREDA F J, GIRON-SIERRA J M, et al. UAV attitude estimation using unscented kalman filter and TRIAD[J]. IEEE Transactions on Industrial Electronics,2012,59(11):4465-4474.
[2] 孙晓哲,宋晗,陈宗基,等. 民机航空电子系统及虚拟样机技术研究[C]//大型飞机关键技术高层论坛暨中国航空学会2007年年会论文集,2007:591-597.
[3] 张小林,赵宇博,范力思,等. 新一代高性能无人机飞控系统的研究与设计[J]. 计算机测量与控制,2010,18(11):2588-2590,2601.
[4] 刘金硕,江庄毅,徐亚渤,等. PMVS算法的CPU多线程和GPU两级粒度并行策略[J]. 计算机科学,2017,44(2):296-301.
[5] 胡伟武,李国杰. 纳米级工艺对微处理器设计的挑战[J]. 中国集成电路,2008,17(7):10-24.
[6] 龙达峰,刘俊,张晓明,等. 高速旋转弹飞行姿态磁测解算方法[J]. 弹道学报,2013,25(2):69-73,84.
[7] 邢丽,熊智,刘建业,等. 针对高动态载体应用的高精度捷联惯导姿态算法优化方法[J]. 中国惯性技术学报,2014, 22(6):701-706.
[8] 王天宇,张晓明,关洋,等. 基于DMA的常规弹导航信息实时处理技术[J]. 传感器与微系统,2015,34(12):25-28.
[9] CONNER M. Serial bus enhances venerable I2C-bus speed[J]. EDN,2006(11):26.
[10] STMicroelectronics. Getting started with STM32F4xxxx MCU hardware development[Z].AN4488.2015.
[11] 袁政. 无人机航姿参考系统开发及信息融合算法研究[D].福州:中南大学,2012.
[12] CODY W J, WAITE W M. Software manual for the elementary functions[M]. Prentice-Hall,1980:102-110.
[13] 于敦山,沈绪榜. 反规格化浮点数除法的一种实现方法[C]//第五届计算机工程与工艺学术年会论文集,1997:1-5.
[14] DEKKER T J, HOFFMANN W. Rehabilitation of the gauss-jordan algorithm[J]. Numerische Mathematik,1989, 54(5):591-599.
[15] LIU R, ZHAO H S, LI X J. Design of 32-bit floating-point multiplier based on improved carry-select adder[J]. Modern Electronics Technique,2013,36(16):133-136.
[16] STMicroelectronics. Floating point unit demonstration on STM32 microcontrollers[Z]. AN4044,2016.
[17] STMicroelectronics. Managing memory protection unit (MPU) in STM32 MCUs[Z]. AN4838,2016.
