城轨车-地场景下基于CGAN-LSTM网络的OTFS-ISAC系统信道估计  

OTFS-ISAC system channel estimation based on GAN-LSTM network in urban rail train-infrastructure scenario

作　　者：杨骞[1,2] 苏宏升 陶旺林[1,3] 刘大为 YANG Qian;SU Hongsheng;TAO Wanglin;LIU Dawei(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;School of Electrical Engineering,Lanzhou Institute of Technology,Lanzhou 730300,China;China Mobile Communications Group Gansu Company Limited,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070 [2]兰州工业学院电气工程学院,甘肃兰州730300 [3]中国移动通信集团甘肃有限公司,甘肃兰州730070

出　　处：《通信学报》2025年第2期59-71,共13页Journal on Communications

基　　金：甘肃省高校教师创新基金资助项目(No.2025B-239);兰州工业学院青年科技创新基金资助项目(No.2024KJ-16);甘肃省高校青年博士支持项目(No.2023QB-049)。

摘　　要：为解决商用B5G/6G城轨车-地场景下通信感知一体化(ISAC)信号传输信道估计问题,提出了一种基于深度学习的信道估计方法。建立基于正交时频空(OTFS)调制的ISAC信号传输系统模型,引入OTFS导频辅助,设计条件生成对抗网络和长短期记忆网络结合的CGAN-LSTM,将混沌博弈优化算法与经典Adam优化器结合,对网络参数进行优化,利用优化网络完成信道估计。仿真结果表明,所提方法在归一化均方误差和误码率方面,优于传统的信道估计方法,为ISAC信号检测和恢复提供必要数据基础。In order to solve the problem of integrated sensing and communication(ISAC)signal transmission channel estimation in commercial B5G/6G urban rail train-infrastructure scenario,a channel estimation method based on deep learning was proposed.An ISAC signal transmission system model based on orthogonal time frequency space(OTFS)modulation was established,the OTFS pilot was introduced,with OTFS pilot introduced to aid,CGAN-LSTM combining conditional generative adversarial network(CGAN)and long short-term memory(LSTM)network was designed.Chaos game optimization(CGO)algorithm was combined with classical Adam optimizer to optimize the network parameters,and the optimized network was used to complete the channel estimation.Simulation results show that the proposed method is superior to traditional channel estimation methods in normalized mean square error and bit error rate,and provides necessary data basis for ISAC signal detection and recovery.

关 键 词：通信感知一体化 正交时频空 条件生成对抗网络 长短期记忆 混沌博弈优化 

分 类 号：TN92[电子电信—通信与信息系统]