To address the need for massive connections in Internet-of-Vehicle communications, local wireless networks utilize non-orthogonal multiple access (NOMA). Scholars have introduced deep reinforcement learning networks for user grouping and power allocation to reduce computational complexity. However, the traditional algorithm based on DQN (Deep Q-Network) still exhibits slow convergence speed and low training stability, while the uniform sampling method in the sample playback process suffers from low sampling efficiency. In order to address these issues, this paper proposes a user grouping and power allocation method for NOMA systems based on Prioritized Dueling DQN-DDPG joint optimization. Firstly, the paper introduces the user grouping network based on Dueling DQN, which considers both the state value and action value in the entire connection layer. The two values compete with each other, are summed up, and re-evaluated. The network significantly improves training stability and increases the convergence speed. Secondly, in this paper, a depth deterministic strategy gradient (DDPG) algorithm with symmetric properties is used. This algorithm works well for continuous action spaces and avoids the power quantization error because of the continuity of power value in the power allocation stage. Finally, the priority sampling based on TD-error (Temporal-difference error) is combined with the Dueling DQN network and DDPG network to ensure random sampling and improve the replay probability of important samples. Simulation results show that the proposed priority-based Dueling DQN-DDPG algorithm significantly improves the convergence speed of sample training. The research results of this paper provide a solid foundation for the following research content, which focuses on NOMA system resource allocation under the mobile user state.

To meet the demands of massive connections in the Internet-of-vehicle communications, non-orthogonal multiple access (NOMA) is utilized in the local wireless networks. In NOMA technique, various optimization methods have been proposed to provide optimal resource allocation, but they are limited by computational complexity. Recently, the deep reinforcement learning network is utilized for resource optimization in NOMA system, where a uniform sampled experience replay algorithm is used to reduce the correlation between samples. However, the uniform sampling ignores the importance of sample. To this point, this paper proposes a joint prioritized DQN user grouping and DDPG power allocation algorithm to maximize the system sum rate. At the user grouping stage, a prioritized sampling method based on TD-error (temporal-difference error) is proposed. At the power allocation stage, to deal with the problem that DQN cannot process continuous tasks and needs to quantify power into discrete form, a DDPG network is utilized. Simulation results show that the proposed algorithm with prioritized sampling can increase the learning rate and perform a more stable training process. Compared with the previous DQN algorithm, the proposed method improves the sum rate of the system by 2% and reaches 94% and 93% of the exhaustive search algorithm and optimal iterative power optimization algorithm, respectively. Although the sum rate is improved by only 2%, the computational complexity is reduced by 43% and 64% compared to the exhaustive search algorithm and the optimal iterative power optimization algorithm, respectively.

In indoor low-light environments, the lack of light makes the captured images often suffer from quality degradation problems, including missing features in dark areas, noise interference, low brightness, and low contrast. Therefore, the feature extraction algorithms are unable to extract the feature information contained in the images accurately, thereby hindering the subsequent target search task in this environment and making it difficult to determine the location information of the target. Aiming at this problem, a joint local and high-level semantic information (JLHS) target search method is proposed based on joint bilateral filtering and camera response model (JBCRM) image preprocessing enhancement. The JBCRM method improves the image quality by highlighting the dark region features and removing the noise interference in order to solve the problem of the difficult extraction of feature points in low-light images, thus providing better visual data for subsequent target search tasks. The JLHS method increases the feature matching accuracy between the target image and the offline database image by combining local and high-level semantic information to characterize the image content, thereby boosting the accuracy of the target search. Experiments show that, compared with the existing image-enhancement methods, the PSNR of the JBCRM method is increased by 34.24% at the highest and 2.61% at the lowest. The SSIM increased by 63.64% at most and increased by 12.50% at least. The Laplacian operator increased by 54.47% at most and 3.49% at least. When the mainstream feature extraction techniques, SIFT, ORB, AKAZE, and BRISK, are utilized, the number of feature points in the JBCRM-enhanced images are improved by a minimum of 20.51% and a maximum of 303.44% over the original low-light images. Compared with other target search methods, the average search error of the JLHS method is only 9.8 cm, which is 91.90% lower than the histogram-based search method. Meanwhile, the average search error is reduced by 18.33% compared to the VGG16-based target search method. As a result, the method proposed in this paper significantly improves the accuracy of the target search in low-light environments, thus broadening the application scenarios of target search in indoor environments, and providing an effective solution for accurately determining the location of the target in geospatial space.

As the number of mobile users and video traffics grow explosively, the data rate demands increase tremendously. To improve the spectral efficiency, the spectrum are reused inter cell or intra cell, such as the ultra dense network with multi-cell or the cellular network with Device-to-Device communications, where the co-channel interferences are brought and needs to be suppressed. According to the time-frequency energy sensing to the communication signals, the desired signal and the interference signal have different energy concentration areas on the time frequency plane, which provide opportunities to suppress the co-channel interference with time varying filter. This paper analyzes the time-frequency distributions of the Gaussian pulse shaping signals, discusses the effect of the analyzing window length on the time-frequency resolution, exploits the equivalence between the time frequency analysis at the baseband and at the radio front end, and finally reveals the advantages of the proposed masking threshold constrained time varying filter based co-channel interference mitigation method. The pass region for the linear time varying filter is generated according to the time-varying energy characteristics of the I/Q separated 4-QAM pulse shaping signals, where the optimum masking threshold is obtained by the optimum-BER criterion. The proposed co-channel interference suppression method is evaluated in aspect of BER performance, and simulation results show that the proposed method outperforms existing methods with low-pass or band-pass filters.

In the mobile communication system, non-orthogonal multiple access (NOMA) technology is introduced to improve spectrum efficiency. Because the combination of users and the transmission power of each user are very important to the performance of NOMA system, the resource allocation technology of NOMA system has been widely studied. In recent years, scholars have introduced deep reinforcement learning network for user grouping and power allocation, which can effectively reduce the computational complexity and improve the system sum rate. However, the traditional algorithm based on DQN network still has the problems of slow convergence speed and low training stability, and the uniform sampling method in the sample playback process has the problem of low sampling efficiency. To address these problems, this paper proposes a priority-based user grouping and power allocation method of NOMA system optimized by Dueling DQN-DDPG, which can effectively improve the convergence speed and training stability. Firstly, in the user grouping stage, a user grouping network based on Dueling DQN is proposed. This network considers both the state value and the action value in the whole connection layer. The two values compete with each other, and then they are summed up and re-evaluated. The proposed network can effectively improve the stability of the traditional DQN network training process and speed up the training convergence. Secondly, considering the continuity of power value, DDPG network, which is suitable for dealing with continuous action space, is adopted in the power allocation stage, which can avoid the power quantization error. Finally, the priority sampling based on TD-error is combined with Dueling DQN network and DDPG network respectively, which can ensure random sampling and improve the replay probability of important samples. Simulation results show that the priority based Dueling DQN -DDPG algorithm proposed in this paper can greatly improve the convergence speed of sample training. At the same time, this scheme has the advantage of priority sampling, which can improve the learning speed and make the training process more stable. Compared with the traditional DQN algorithm, the convergence speed of the proposed algorithm is nearly doubled, and the training process is more stable, but the computational complexity is only increased by about 15%.