Unmanned aerial vehicles (UAVs) are regarded as a promising solution for mobile-edge computing (MEC) systems due to their flexibility and capability to provide computing services to ground terminals (GTs). By leveraging UAVs, the latency in computation tasks can be reduced significantly, particularly in disaster scenarios. Additionally, reconfigurable intelligent surfaces (RISs) have emerged as a novel technology for enhancing the wireless propagation environment in wireless networks. This article proposes a multi-UAV-assisted MEC system where computation tasks of GTs can be computed locally or partially offloaded to UAVs. Furthermore, practical RIS phase shift designs are considered to enhance the communication performance between GTs and UAVs. To minimize the system delay and achieve fairness among GTs, the computation offloading strategy, trajectory of the UAVs are optimized using a Markov decision process. Simultaneously, the RIS phase shift is optimized through an alternating optimization algorithm. Additionally, a cooperative multiagent deep reinforcement learning framework is developed to obtain a optimal solution by employing the multiagent twin delayed deep deterministic policy gradient (MATD3) algorithm. Numerical results indicate that MATD3 can effectively improve the system delay and fairness performance of the RIS-assisted multi-UAV MEC system, as compared to benchmark solutions.

To face the challenges in emergency scenarios, a multi-task and multi-objective optimization algorithm for computation offloading and relay communication is investigated for the air-ground integrated networks, composed of unmanned aerial vehicles (UAVs), emergency vehicle users (EVUs) and ground sensor nodes (GSNs). We propose an HFL-DDQN algorithm, which combines horizontal federated learning (HFL) with double deep Q-network (DDQN). First, UAVs are separated into two clusters according to the services they provide, i.e., edge computing or relay communication. Next, the optimization problems are formulated for two types of services respectively. For the computation offloading tasks of EVUs, the optimization objective is to minimize the weighted sum of delay and energy consumption. For the sensor data transmission of GSNs, the optimization objective is to maximize the minimum rate of relay links. We define the total cost of the system as the sum of two types of services. Then, federated aggregation is used to joint training the global neural networks model without sharing raw data. Furthermore, the DDQN is improved by adopting prioritized experience replay to achieve better convergence. The simulation results show that the proposed HFL-DDQN algorithm not only outperforms the state-of-art baselines in terms of the system cost, but also promotes the generalization in execution process, which is especially applicable to the rescue scene under accidents.

In the evolution of the Internet of vehicles (IoV), the increasing demand for vehicular computation tasks presents significant challenges, particularly in the context of constrained local computation resources and high processing delays. To mitigate these challenges, multi-access edge computing (MEC) offers a potential solution by leveraging edge servers for low-latency processing. However, it also encounters issues such as sub-channel competition and workload imbalance owing to the uneven distribution of vehicle densities. This paper introduces a novel IoV architecture that incorporates multi-task and multi-roadside unit (RSU) capabilities, enabling edge-to-edge collaboration for efficient task offloading among RSUs. The optimization problem is formulated with the objective of minimizing the overall task delay, which is further divided into two sub-problems: communication resource allocation and load balancing. Considering the non-deterministic polynomial (NP)-hard nature of these sub-problems, we propose a two-stage deep reinforcement learning-based communication resource allocation and load balancing (DRLCL) algorithm to address them sequentially. Based on realistic vehicle trajectories, comprehensive evaluation results demonstrate the superiority of the proposed algorithm in reducing system delay compared to existing state-of-the-art baselines, offering an effective approach for optimizing the performance of vehicular edge computing (VEC) networks.

In this paper, we propose a distributed sidelink resource allocation approach named federated multi-agent deep Q network (FMDQN). The objective is the network spectrum-energy efficiency (SEE) maximization. To address the optimization problem in a distributed manner, we merge federated learning and deep reinforcement learning. We establish an asynchronous federated learning framework by facilitating local model training, thus avoiding the transmission of user-sensitive raw data. Our proposed approach utilizes a client-server architecture and federated aggregation. Additionally, we formulate a sub-band and power allocation strategy that aims to maximize SEE while adhering to strict delay limits for the given task. Simulation results illustrate the favorable convergence of the proposed algorithm in highly dynamic and uncertain Internet of vehicles (IoV) environments.

New radio vehicle-to-everything (NR-V2X) is an emerging technology based on the 5G cellular mobile communication networks. In this paper, the sidelink (SL) resource allocation for NR-V2X in Mode 1 is considered, where 5G base station (gNB) schedules SL resources to be used by V2X users (VUs) based on the periodically reported channel state information (CSI). To reduce the overhead, an energy efficient power allocation scheme of CSI transmission is proposed using hybrid spectrum access technology. With the knowledge of CSI, SL resource allocation is modeled as a mixed binary integer nonlinear programming (MBINP). It is designed to maximize the sum throughput of NR-V2X networks among different subcarriers subject to the total available power and the minimal transmission rate. To achieve that, the original MBINP is decomposed into two sub-problems, namely, subcarrier assignment and power allocation. Firstly, the appropriate subcarriers are quickly obtained by averaging the power allocation. Secondly, an alternative optimization mechanism is proposed to the power allocation. Simulation results show that the proposed power allocation scheme of CSI transmission can obviously save the energy consumption, and the proposed suboptimal SL resource allocation algorithm achieves better performance than the compared methods with relatively low complexity.