• Study On Optical Communications
  • Vol. 49, Issue 3, 71 (2023)
Ming-hu WU, Bo JIN, Nan ZHAO*, and Ru WANG
Author Affiliations
  • School of Electrical and Electronic Engineering, Hubei University of Technology, Wuhan 430068, China
  • show less
    DOI: 10.13756/j.gtxyj.2023.03.012 Cite this Article
    Ming-hu WU, Bo JIN, Nan ZHAO, Ru WANG. Spectrum Resourse Management Method of V2X based on Deep Reinforcement Learning[J]. Study On Optical Communications, 2023, 49(3): 71 Copy Citation Text show less
    V2X communication system model of single V2V link
    Fig. 1. V2X communication system model of single V2V link
    The model diagram of Dueling-DQN algorithm
    Fig. 2. The model diagram of Dueling-DQN algorithm
    The efficiency of V2I linkη1 on different transmission load K
    Fig. 3. The efficiency of V2I linkη1 on different transmission load K
    The efficiency of V2V linkη2 on different transmission load K
    Fig. 4. The efficiency of V2V linkη2 on different transmission load K
    The efficiency of V2X communication networkη3 on different transmission load K
    Fig. 5. The efficiency of V2X communication networkη3 on different transmission load K
    The V2I link efficiency η1 of three algorithms at different speeds
    Fig. 6. The V2I link efficiency η1 of three algorithms at different speeds
    The V2V link efficiency η2 of three algorithms at different speeds
    Fig. 7. The V2V link efficiency η2 of three algorithms at different speeds
    The V2X communication network efficiency η3 of three algorithms at different speeds
    Fig. 8. The V2X communication network efficiency η3 of three algorithms at different speeds
    参数类型数值
    V2I链路数量4
    V2V链路数量1
    载波频率/GHz2
    总带宽W/MHz10
    V2I功率等级/dBm23
    V2V功率等级/dBm[25,15,10,0]
    电路损耗pc/dBm16
    基站天线增益/dBi8
    车辆天线增益/dBi3
    噪声功率σ2/dBm-114
    传输负载K/Bytes2×1 060
    车辆速度V/km/h30
    Table 1. Vehicle environment simulation parameters
    参数V2I链路V2V链路
    路径损耗模型128.1+37.6 log10dLOSin WINNER+B1 Manhattan
    阴影分布对数正态分布对数正态分布
    阴影标准差/dB83
    去相关距离/m5010
    路径损耗和阴影更新/ms100100
    快速衰落瑞利衰落瑞利衰落
    快速衰落更新/ms11
    Table 2. Channel model parameters of V2I link and V2V link
    参数类型数值
    隐藏层数量3
    学习率0.001
    优化器RMSProp
    更新步数100
    激活函数ReLu
    小批样本数2 000
    折扣因子0.5
    Table 3. The hyperparameters of Dueling-DQN
    Ming-hu WU, Bo JIN, Nan ZHAO, Ru WANG. Spectrum Resourse Management Method of V2X based on Deep Reinforcement Learning[J]. Study On Optical Communications, 2023, 49(3): 71
    Download Citation