基于深度学习的测控数据传输技术  

TT&C Data Transmission Technology Based on Deep Learning

作　　者：丁丹 杨柳 刘步花 DING Dan;YANG Liu;LIUBuhua(Department of Electronic and Optical Engineering,Aerospace Engineering University,Beijing 101416,China)

机构地区：[1]航天工程大学电子与光学工程系,北京101416

出　　处：《无线电工程》2020年第4期313-317,共5页Radio Engineering

摘　　要：传统测控数据传输根据试验场区环境和预设指标进行各模块的分立设计和局部优化,难以对战时复杂、时变信道函数进行全局最优拟合,严重影响战时系统整体性能的发挥。针对此问题,研究基于深度学习的测控数据传输技术,运用深度学习网络代替传统调制解调、信道编解码和信道均衡等多个分立模块,利用神经元的灵活组合逼近复杂的函数,将系统调节为全局最优。仿真结果表明,在“非线性+多径”信道条件下,该方法能够将传输容量提升至传统体制2倍以上,或将链路余量提高3 dB以上,同时保持与传统测控数据传输体制相当的系统复杂度。Traditional TT&C data transmission,designed module by module and optimized locally according to test site’s environment and preset indexes,is not able to optimally fit the complex and time variant channel function,resulting in serious performance degradation of the system during the war.To overcome this problem,the TT&C data transmission based on deep learning is studied,in which the deep learning network replaces traditional separate modules such as modulation and demodulation,channel coding and decoding,and channel equalization and the complex function is approximated by neurons’agile combination,so that the whole system is adjusted to global optimization.Simulation results show that under complex channel of“nonlinear plus multipath”,the transmission capacity of this method will be 2 times or more larger than that of the traditional one,or the link margin could be increased by 3 dB or more,and at the same time it shows similar system complexity as the traditional TT&C data transmission system.

关 键 词：测控系统 数据传输 深度学习 神经网络 传输性能 

分 类 号：V249.3[航空宇航科学与技术—飞行器设计;航空宇航科学技术]