In the low-altitude three-dimensional coverage scenario with unmanned aerial vehicles (UAVs), user data requirements change quickly and the asymmetry of uplink and downlink traffic is hard to address. We utilize cell-free radio access networks (CF-RAN) to support dynamic scenario due to its cooperative capability and scalability, and adopt network-assisted full-duplex (NAFD) to support flexible duplex communication by selecting appropriate node duplex modes according to the traffic in real-time. Considering the scalable minimum mean square error (MMSE) receiver and regularized zero-forcing (RZF) precoding scheme, the closed-form expressions of uplink spectral efficiency with infinite block-length regime and the lower bound of downlink spectral efficiency with finite block-length communication (FBLC) regime are derived. Based on these expressions, we propose a multi-objective optimization problem (MOOP) to maximize and balance the spectral efficiency of upink and downlink by optimizing the duplex mode of remote radio units (RRUs) and solving it with deep Q-learning (DQN) algorithm. Simulation results verify the accuracy of derived expressions, the effectiveness of the proposed optimization algorithm and the superiority of NAFD technique in low-altitude CF-RAN system.

Cell-free radio access network (CF-RAN) with network-assisted full-duplex (NAFD) possesses scalability and enables reception and transmission simultaneously, which is suitable for large-scale ultra reliable and low-latency communication (URLLC). Achieving strict requirements of URLLC for each terminal with a fixed infrastructure is challenging, and unmanned aerial vehicles (UAVs) have been considered as promising enablers to handle this issue due to its flexible deployment, low cost and large coverage. In this paper, we investigate a multi-UAV aided CF-RAN with NAFD that use UAVs as aerial access points (APs). We firstly propose a multi-UAV deployment algorithm based on user distribution and quality of service (QoS) requirement. Then, we derive the closed-form expressions for uplink achievable rate with long block length regime and the lower bound of downlink achievable rate with finite block length communication (FBLC) regime. Based on those expressions, we formulate a weighted sum spectral efficiency maximization problem and solve it by Deep Q-Network (DQN) algorithm. Numerical results verify the accuracy of the derived closed-form expressions and illustrate the impact of system parameters on spectral efficiency. The effectiveness of proposed UAV deployment algorithm and the performance advantages of the weighted sum spectral efficiency optimization algorithm are demonstrated.

Cellular-connected unmanned aerial vehicles (UAVs) is considered as integral components for sixth generation (6G) cellular networks. Cell-free radio access network (CF-RAN) with network-assisted fullduplex (NAFD) possesses global collaborative capabilities and enable uplink and downlink transmission simultaneously, which can be considered as a potential technology for supporting air-ground communication. In this paper, we investigate the performance of cellular-connected UAV in CF-RAN with NAFD. We propose a modified beamforming training scheme to mitigate cross-link interference (CLI) and a location-aware access point (AP) clustering strategy to reduce fronthaul overhead. We derive closed-form expressions for uplink and downlink achievable rates of GUEs and UAVs, respectively. Based on these expressions, we propose an efficient global spectral efficiency optimization scheme by solving a multi-objective optimization problem (MOOP) aiming to maximize the uplink sum rates and downlink sum rates simultaneously with deep Q-network (DQN). Numerical results verify the accuracy of the derived closed-form expressions. The effectiveness of the modified beamforming training scheme and location-aware AP clustering strategy are proved. In addition, the impact of system parameters and the advantages of NAFD system are analyzed. We also illustrate the convergence and benefits of DQN-based optimization scheme on different type of users.

There have been increasing interests in graph pattern mining due to the booming of data volume in various domains. Conventional graph mining implementations which calculate the exact count of patterns usually suffer from huge amounts of intermediate data and low performance on large-scale graphs. With the observation that the exact pattern counts are not required in many real-world graph pattern mining problems, previous works (e.g., ASAP) proposed an approximate graph pattern mining algorithm and improved the performance of graph pattern mining by up to two orders of magnitudes. The crucial sampling operation in the ASAP algorithm exposes high parallelism and complex edge searching. Moreover, the performance of ASAP is closely related the sampling order. However, previous works failed to tackle these problems in the design. Thus, we propose a novel Processing-in-Memory (PIM) architecture for parallel approximate graph pattern mining problems. We introduce dictionaries on the logic layer of PIM devices for edge indexing. We also explore the design space of sampling orders and give the optimal sampling strategy. The comprehensive experimental results show that, our design achieves up to 97 times performance improvement against ASAP system.