This paper investigates a dynamic 3D trajectory design of multiple cache-enabled unmanned aerial vehicles (UAVs) in a wireless device-to-device (D2D) caching network with the goal of maximizing the long-term network throughput. By storing popular content at the nearby mobile user devices, D2D caching is an efficient method to improve network throughput and alleviate backhaul burden. With the attractive features of high mobility and flexible deployment, UAVs have recently attracted significant attention as cache-enabled flying base stations. The use of cache-enabled UAVs opens up the possibility of tracking the mobility pattern of the corresponding users and serving them under limited cache storage capacity. However, it is challenging to determine the optimal UAV trajectory due to the dynamic environment with frequently changing network topology and the coexistence of aerial and terrestrial caching nodes. In response, we propose a novel multi-agent reinforcement learning based framework to determine the optimal 3D trajectory of each UAV in a distributed manner without a central coordinator. In the proposed method, multiple UAVs can cooperatively make flight decisions by sharing the gained experiences within a certain proximity to each other. Simulation results reveal that our algorithm outperforms the traditional single- and multi-agent Q-learning algorithms. This work confirms the feasibility and effectiveness of cache-enabled UAVs which serve as an important complement to terrestrial D2D caching nodes.

One of the key applications in the fifth generation (5G) cellular networks is to support ultra-reliable and lowlatency communication (URLLC) which requires extremely high reliability ( 99:999%) and low latency (< 1 ms). In this paper, the energy efficiency (EE) is maximized for downlink multiuser multiple-input single-output (MISO) networks under short packet transmission. We jointly optimize the precoders at the base station (BS) for serving multiple downlink users and the decoding error probability (DEP) with finite blocklength (FBL) codes, subject to the constraints on the BS transmit power and DEP per URLLC user. This non-convex optimization problem is then accurately approximated by applying the Dinkelbach method and approximating the channel dispersion in the high signal-to-interference-plus-noise ratio (SINR) regime and the entire (practical) SINR regime, respectively. The resulting problems are convex in the design variables of the precoders and DEPs when the remaining variables are held fixed, and using this fact we provide iterative algorithms to efficiently find locally optimal solutions. Furthermore, the convergence of the proposed algorithms and the closed-form expressions of the precoders are also derived. Simulation results validate the effectiveness of the proposed algorithms that support energy-efficient URLLC as well as the impact of various network parameters such as the blocklength, number of users, distance between the BS and URLLC users and the threshold of the DEP on the EE and spectral efficiency (SE) performance.

In this paper, we study an amplify-and-forward (AF) relay network with energy harvesting (EH) source and relay nodes. Both nodes can continuously harvest energy from the environment and store it in batteries with finite capacity. Additionally, the source node is capable of transferring a portion of its energy to the relay node through a dedicated channel. The network performance depends on not only the energy arrival profiles at EH nodes but also the energy cooperation between them. We jointly design power control and transfer for maximizing the sum rate over finite time duration, subject to energy causality and battery storage constraints. By introducing auxiliary variables to confine the accumulated power expenditure, this non-convex problem is solved via a successive convex approximation (SCA) approach, and the local optimum solutions are obtained through dual decomposition. Also when channels are quasi-static and the power control values of the source (relay) node are preset to a constant, a monotonically increasing power control structure with the time is revealed for the relay (source) node with infinite battery capacity. Computer simulations are used to validate the theoretical findings and to quantify the impact of various factors such as EH intensity at nodes and relay position on the sum rate performance.

Mobile relaying is envisioned as a promising technology to alleviate the masking effect in satellite links. Nevertheless, due to the mobility and limited communication range, the network topology in mobile relay networks becomes highly dynamic and time-varying, resulting in increased difficulty of network optimization. This letter investigates unmanned aerial vehicles (UAVs) aided hybrid satellite-terrestrial network, where UAVs are served as mobile relay base stations to assist the communication between the satellite and ground users. We aim to maximize the number of served users by optimizing UAV trajectory and user link selection. To solve the formulated mixed-integer and non-convex optimization problem, we first find the optimal link selection via a designed graph neural network (GNN), and then adjust the UAV locations by using model-free reinforcement learning (RL), alternately. Numerical results demonstrate that our proposed scheme is superior to the state-of-the-art RL algorithms.

This study explores a new resource allocation framework for a multi-user multiple-input and multiple-output (MU-MIMO) downlink system utilizing aerial reconfigurable intelligent surfaces (ARIS) in a rate-splitting multiple access (RSMA) scheme under ultra-reliable low-latency communication (URLLC). The objective is to maximize the overall spectral efficiency of users located in severely blocked dead zones, where there is no direct link between the base station (BS) and the users and, the ARIS acts as a relay to facilitate communication. To achieve this, we formulate a sum-rate maximization problem considering the joint design of active beamforming at the BS, passive beamforming at the ARIS, and determination of the ARIS height. The constraints include given packet size and packet error probability (PEP) requirements. Since the original problem is non-convex, we propose an alternating optimization framework to solve it. The framework involves solving sub-problems of beamforming design and ARIS location alternately. The beamforming design sub-problem is addressed using successive convex approximation, while the ARIS location sub-problem is solved using a particle-swarm optimization (PSO) algorithm. The proposed solution demonstrates fast convergence within a few iterations, as validated through numerical simulations. Additionally, a computational complexity analysis is performed. The results show that the proposed RSMA scheme for the ARIS-aided DL URLLC system surpasses conventional multiple access schemes in terms of rate performance. The findings highlight the potential of utilizing ARIS and RSMA to enhance the spectral efficiency and meet the requirements of URLLC applications.

In this article, we consider multiple solar-powered wireless nodes (WNs) which utilize the harvested solar energy to transmit collected data to multiple unmanned aerial vehicles (UAVs) in the uplink. In this context, we jointly design UAV flight trajectories, UAV-node user association, and uplink power control to effectively utilize the harvested energy and manage co-channel interference within a finite time horizon. The design goal is to ensure the fairness of WNs by maximizing the worst user rate. The joint design problem is highly nonconvex and requires causal (future) knowledge of the instantaneous energy state information (ESI) and channel state information (CSI), which are difficult to predict in reality. To overcome these challenges, we propose an offline method based on convex optimization that only utilizes the average ESI and CSI, where line-of-sight (LOS) and non-LOS (NLOS) channels are considered. The problem is solved by three convex subproblems with successive convex approximation (SCA) and alternative optimization. We further design an online convex-assisted reinforcement learning (CARL) method based on real-time environmental information. An idea of multi-UAV regulated flight corridors, based on the optimal offline UAV trajectories, is proposed to avoid unnecessary flight exploration by UAVs and enables us to improve the learning efficiency and system performance, as compared with the conventional reinforcement learning (RL) method. Computer simulations are used to verify the effectiveness of the proposed methods. The proposed CARL method provides 25% and 12% improvement on the worst user rate over the offline and conventional RL methods.