基于LSTM的TTE网络速率约束流量预测

史亚菲; 李峭; 熊华钢

doi:10.13700/j.bh.1001-5965.2019.0320

基于LSTM的TTE网络速率约束流量预测

doi: 10.13700/j.bh.1001-5965.2019.0320

北京航空航天大学电子信息工程学院, 北京 100083

基金项目:

国防科技基金 0101070

中央高校基本科研业务费专项资金 YWF-14-DZXY-018

载人航天预先研究项目 060301

详细信息

作者简介:
史亚菲女, 硕士研究生。主要研究方向:实时网络、航空电子系统综合化互连

李峭男, 博士, 讲师, 硕士生导师。主要研究方向:航空电子网络、分布式实时系统

熊华钢男, 博士, 教授, 博士生导师。主要研究方向:航空电子综合、机载网络

通讯作者:
李峭, E-mail: avionics@buaa.edu.cn

中图分类号: V247;TP393
计量
- 文章访问数: 1053
- HTML全文浏览量: 139
- PDF下载量: 457
- 被引次数: 0
出版历程
- 收稿日期: 2019-06-20
- 录用日期: 2019-07-05
- 网络出版日期: 2020-04-20

Rate-constrained traffic prediction of TTE network based on LSTM

School of Electronic and Information Engineering, Beihang University, Beijing 100083, China

Funds:

National Defense Science and Technology Fund 0101070

the Fundamental Research Funds for the Central Universities YWF-14-DZXY-018

Manned Space Pre-Research Project 060301

More Information

Corresponding author: LI Qiao, E-mail: avionics@buaa.edu.cn

摘要

摘要:
时间触发以太网（TTE）中的速率约束（RC）流量为事件触发流量，在RC流量动态调度的应用场景下，若能预测未来短时间内数条RC流量到达交换节点的序列，使交换节点提前进行调度决策，以减小RC流量时延，提高网络吞吐量。对RC流量到达序列预测问题进行了研究，建立了RC流量的到达序列模型，提出了基于长短期记忆网络（LSTM）算法的RC流量预测算法。利用OMNET++工具进行TTE网络仿真，得到多组混合关键性配置下RC流量的传输数据；以此作为输入样本对预测算法进行训练和测试。实验结果显示，LSTM算法在RC流量预测问题的准确率达到了70%以上。通过对比实验说明所提算法适用于RC流量预测场景。
- 时间触发以太网(TTE)网络 /
- 速率约束(RC)流量 /
- 流量预测 /
- 长短期记忆网络(LSTM)算法 /
- 网络仿真
Abstract:
The rate constraint (RC) traffic in time triggered Ethernet (TTE) is event-triggered traffic. In the application scenario of dynamic scheduling of RC traffic, if it can predict the sequence of several RC traffic arriving at the switching node in a short time in the future, the switching node can make scheduling decision in advance to reduce RC traffic delay and improve network throughput. In this paper, the arrival sequence model of RC traffic is established, and an algorithm of RC traffic prediction based on long-term memory network (LSTM) is proposed. Using OMNET++ tool to simulate TTE network, we can get the data of RC traffic transmission under multiple groups of mixed critical configuration, and train and test the prediction algorithm as an input sample. The experimental results show that the accuracy of LSTM algorithm in RC traffic prediction is more than 70%. The experimental results show that the proposed algorithm is suitable for RC traffic prediction scenarios.
- time-triggered Ethernet (TTE) network /
- rate-constrained (RC) traffic /
- traffic prediction /
- long short-term memory (LSTM) algorithm /
- network simulation

HTML全文

图 1 TTE网络TT、RC流量的传输过程

Figure 1. Transmission process of TT and RC traffic in TTE network

下载: 全尺寸图片幻灯片

图 2 TTE网络流量传输示例

Figure 2. Examples of TTE network traffic transmission

下载: 全尺寸图片幻灯片

图 3 RC流量预测排序效果

Figure 3. Sorting effect of RC traffic sorting

下载: 全尺寸图片幻灯片

图 4 RNN模型

Figure 4. RNN model

下载: 全尺寸图片幻灯片

图 5 LSTM算法模型

Figure 5. LSTM algorithm model

下载: 全尺寸图片幻灯片

图 6 LSTM算法详解

Figure 6. Details of LSTM algorithm

下载: 全尺寸图片幻灯片

图 7 LSTM预测模型数据集构造图

Figure 7. Dataset construction map of LSTM prediction model

下载: 全尺寸图片幻灯片

图 8 网络拓扑结构示意图

Figure 8. Schematic diagram of network topology

下载: 全尺寸图片幻灯片

图 9 算法预测值与真实值对比

Figure 9. Comparison of predicted value and real value of algorithms

下载: 全尺寸图片幻灯片

表 1 TTE网络流量配置

Table 1. TTE network traffic configuration

消息类型	源节点	目的节点	数据帧长/Byte	周期/ms
TT1	ES₁	ES₁₂	528	64
TT2	ES₁₁	ES₂₁	215	12
TT3	ES₅	ES₁₁	1505	16
TT4	ES₃	ES₂	1273	84
TT5	ES₂₁	ES₂₁	659	40
RC1	ES₁	ES₂₁	688	2
RC2	ES₄	ES₁₂	808	16
RC3	ES₂	ES₂₁	433	4
RC4	ES₃	ES₁₁	433	16
RC5	ES₅	ES₂₁	1518	4
RC6	ES₁	ES₄	1022	8
RC7	ES₃	ES₂₁	208	32
RC8	ES₁₁	ES₂₁	574	2
RC9	ES₃	ES₂₁	828	4

下载: 导出CSV

表 2 RC流量传输数据

Table 2. Transmission data of RC traffic

到达时间/ms	流量编号	包帧长/Byte
3.600852	RC1	688
4.78901	RC3	433
5.600852	RC1	688
7.600852	RC1	688
8.78901	RC3	433
9.600852	RC1	688
10.391926	RC6	1022
11.600852	RC1	688
12.78901	RC3	433
13.600852	RC1	688
15.600852	RC1	688
16.78901	RC3	433
19.4391926	RC6	1022
		
203.553353	RC3	433
207.553353	RC3	433
210.391926	RC6	1022
211.553353	RC3	433
		

下载: 导出CSV

表 3 算法预测结果比较

Table 3. Comparison of algorithm prediction results

算法模型	^*5%准确率/%	^*8%准确率/%	均方误差
LSTM	75.3	93.2	4.5
RNN	42.8	47.2	40.4
DNN	33.9	40.8	53
注：^5%准确率指预测值误差小于RC流量约束最长帧的5%;^8%准确率指预测值误差小于RC流量约束最长帧的8%。

下载: 导出CSV

参考文献(22)

[1]	熊华钢, 王中华.先进航空电子综合技术[M].北京:国防工业出版社, 2009:2-13. XIONG H G, WANG Z H.Advanced avionics integrated technology[M].Beijing:National Defense Industry Press, 2009:2-13(in Chinese).
[2]	WOLFIG R, JAKOVLJEVIC M.Distributed IMA and DO-297: Architectural, communication and certification attributes[C]//Proceedings Digital Avionics Systems Conference.Piscataway, NJ: IEEE Press, 2008: 1.E.4-1-1.E.4-10.
[3]	SAE Aerospace. Time-triggered Ethernet: AS6802^TM[S].Englewood, CO: SAE International, 2011.
[4]	STENER W, BAUER G, HALL B, et al.TTEthernet dataflow concept[C]//8th IEEE International Symposium on Network Computing and Applications.Piscataway, NJ: IEEE Press, 2009: 319-322.
[5]	熊华钢, 李峭, 黄永葵.航空电子全双工交换式以太网标准研究(续)[J].航空标准化与质量, 2008(2):20-23. http://d.old.wanfangdata.com.cn/Periodical/hkbzhyzl200802008 XIONG H G, LI Q, HUANG Y K.Research on avionics full duplex switched Ethernet standards (continued)[J].Aviation Standardization and Quality, 2008(2):20-23(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/hkbzhyzl200802008
[6]	ARINC 664P7-1. Aircraft data network, Part 7: Avionics full duplex switched ethernet (AFDX) network: ARINC 664P7-2005[S].Rio Rancho, NM: ANRIC Standards, 2005.
[7]	尹伟, 缪万胜, 王念伟, 等.大规模复杂系统的开放式软件架构研究[J].航空电子技术, 2017, 48(2):23-29. http://d.old.wanfangdata.com.cn/Periodical/hkdzjs201702005 YIN W, MIU W S, WANG N W, et al.Open software architecture for large-scale complex systems[J].Avionics Technology, 2017, 48(2):23-29(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/hkdzjs201702005
[8]	SANG A, LI S.A predictability analysis of network traffic[J].Computer Networks, 2002, 39(4):329-345. https://www.sciencedirect.com/science/article/abs/pii/S1389128601003048
[9]	LEVIN M, TSAO Y D.On forecasting freeway occupancies and volumes[J].Transportation Research Record Journal of the Transportation Research Board, 1980, 773:47-49. https://trid.trb.org/view/167550
[10]	张九跃, 焦玉栋.基于RBF神经网络的短时交通流量预测[J].山东交通学院学报, 2008, 16(3):32-36. http://d.old.wanfangdata.com.cn/Periodical/sdjtxyxb200803008 ZHANG J Y, JIAO Y D.Short term traffic flow prediction based on RBF neural network[J].Journal of Shandong Jiaotong University, 2008, 16(3):32-36(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/sdjtxyxb200803008
[11]	赵岩, 何鹏.网络流量的非线性组合预测模型应用研究[J].计算机仿真, 2012, 29(6):140-144. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=jsjfz201206036 ZHAO Y, HE P.Application of nonlinear combination forecasting model for network traffic[J].Computer Simulation, 2012, 29(6):140-144(in Chinese). http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=jsjfz201206036
[12]	肖甫, 赵帅帅, 王少辉, 等.基于智能优化的分布式网络流量预测方法[J].北京邮电大学学报, 2015, 38(S1):45-48. http://d.old.wanfangdata.com.cn/Periodical/bjyddx2015z1011 XIAO F, ZHAO S S, WANG S H, et al.Distributed network traffic prediction method based on intelligent optimization[J].Journal of Beijing University of Posts and Telecommunications, 2015, 38(S1):45-48(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/bjyddx2015z1011
[13]	赵季红, 王明欣, 曲桦, 等.一种基于自适应KLMS的卫星网络流量预测算法[J].北京邮电大学学报, 2018, 41(3):51-55. http://d.old.wanfangdata.com.cn/Periodical/bjyddx201803009 ZHAO J H, WANG M X, QU H, et al.A satellite network traffic prediction algorithm based on adaptive KLMS[J].Journal of Beijing University of Posts and Telecommunications, 2018, 41(3):51-55(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/bjyddx201803009
[14]	焦文喆, 翟正军, 王国庆.时间触发AFDX调度设计及实时性分析[J].计算机工程, 2016, 42(7):42-48. http://d.old.wanfangdata.com.cn/Periodical/jsjgc201607007 JIAO W Z, ZHAI Z J, WANG G Q.Design and real-time analysis of time-triggered AFDX scheduling[J].Computer Engineering, 2016, 42(7):42-48(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/jsjgc201607007
[15]	黄婷婷, 余磊.SDAE-LSTM模型在金融时间序列预测中的应用[J].计算机工程与应用, 2019, 55(1):142-148. http://d.old.wanfangdata.com.cn/Periodical/jsjgcyyy201901022 HUANG T T, YU L.Application of SDAE-LSTM model in financial time series forecasting[J].Computer Engineering and Application, 2019, 55(1):142-148(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/jsjgcyyy201901022
[16]	周瑞, 魏正英, 张育斌, 等.基于LSTM递归神经网络的番茄目标产量时间序列预测[J].节水灌溉, 2018(8):66-70. http://d.old.wanfangdata.com.cn/Periodical/jsgg201808014 ZHOU R, WEI Z Y, ZHANG Y B, et al.Time series forecast of tomato target yield based on LSTM recursive neural network[J].Water-saving Irrigation, 2018(8):66-70(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/jsgg201808014
[17]	TAMASSELICEAN D, POP P, STEINER W.Timing analysis of rate constrained traffic for the TTEthernet communication protocol[C]//IEEE International Symposium on Real-Time Distributed Computing.Piscataway, NJ: IEEE Press, 2015: 119-126.
[18]	BENGIO Y, GOODFELLOW I.深度学习[M].北京:人民邮电出版社, 2017. BENGIO Y, GOODFELLOW I.Deep learning[M].Beijing:Posts and Telecom Press, 2017(in Chinese).
[19]	张文刚.基于深度学习的交通预测技术及其在通信中的应用研究[D].成都: 西南交通大学, 2018. http://cdmd.cnki.com.cn/Article/CDMD-10613-1018709289.htm ZHANG W G.Traffic forecasting technology based on in-depth learning and its application in communication[D].Chengdu: Southwest Jiaotong University, 2018(in Chinese). http://cdmd.cnki.com.cn/Article/CDMD-10613-1018709289.htm
[20]	GERS F, SCHMIDHUBER J, CUMMINS F.Learning to forget:Continual prediction with LSTM[J].Neural Computation, 2000, 12(10):2451-2471. doi: 10.1162/089976600300015015
[21]	周德新, 王彦翔.AFDX网络分析及带宽分配间隔设定策略[J].科学技术与工程, 2016, 16(9):226-230. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=kxjsygc201609039 ZHOU D X, WANG Y X.AFDX network analysis and bandwidth allocation interval setting strategy[J].Science and Technology and Engineering, 2016, 16(9):226-230(in Chinese). http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=kxjsygc201609039
[22]	LEÓN A P, PIEDRAHITA E M, SALCEDO O.On-off model applications to communication networks aggregated traffic[J].Cuadernos De Derecho Agrario, 2011, 15(30):129-147.