Research On Resource Allocation Algorithm For NOMA System Restricted By Physical Layer Security

Non-orthogonal multiple access(NOMA)technology can effectively improve the spectrum resource utilization of wireless communication systems by multiplexing signals in the power domain.The issue of wireless physical layer security has recently attracted considerable attention,and once the signals superposed and transmitted in the NOMA system are eavesdropped by illegal users,huge security risks will arise,and therefore the physical layer security issue in NOMA systems is critical.In addition,resource allocation is regarded as an effective method to improve the physical layer security performance of wireless communication systems.With these consideratons,by combining physical layer security with resource allocation in the context of NOMA system,this thesis first establishes an eavesdropping channel models using different user grouping strategies,and then explores how to improve the secrecy sum rate performance through reasonable resource allocation schemes.The main contributions of this thesis are as follows:For the power allocation problem of the NOMA system based on two-user grouping,a sub-carrier power allocation scheme based on deep learning(DL)is proposed.The proposed scheme uses the data processing capability of DL,regards the channel parameters and power allocation results of the full search power allocation(FSPA)as labeled training data of the deep neural network(DNN).These data can be used to optimize the neural network parameters,and finally train a DNN model suitable for the system.If the channel parameters are imported into the trained DNN,the power allocation results can be directly obtained.The simulation results show that through DNN training,the accuracy of the allocation results for the proposed sub-carrier power allocation scheme are as high as 95%;and the same secrecy sum rate is achieved as the FSPA scheme by greatly reducing the complexity.Under the same complexity,the secrecy sum rate of the system is greatly improved compared to the traditional schemes.For the user grouping problem of the NOMA system based on multi-user grouping,a user grouping scheme based on Q-learning is proposed.The proposed scheme uses Q-leraning based on the optimal value idea,interacts with the communication environment,and continuously selects the optimal actions under the established optimization strategy.Through iterative learnings,the user grouping scheme based on Q-learning can dynamically adjust the probability of action selection according to the communication environment,and then choose the best action suitable for the current environment according to the updated probability.Finally,the users are grouped utilizing the final probability of action selection.The simulation results show that the system secrecy sum rate can be rapidly improved after iterative learnings using the proposed user grouping scheme,which prove the effectiveness of the proposed scheme.In addition,compared with the traditional scheme,the proposed scheme can also significantly improve the NOMA system secrecy sum rate.