Fast Learning for Optimization of Green Edge Collaborative UAV

Joint Spectrum, Association, and Deployment Optimization for UAV Swarm-Assisted ISAC Networks

—Unmanned aerial vehicle (UAV) swarm-assisted integrated sensing and communication (ISAC) networks are a crucial technology for providing communication and sensing services in emergency rescue scenarios without base station support. However, the strong coupling between communication and sensing resources in such networks fundamentally limits the communication and sensing performance of ISAC systems. This paper jointly optimizes spectrum allocation, UAV association and deployment to maximize average system throughput while ensuring localization accuracy in such networks, where sensing is realized through localization. We begin by deriving an analytical expression for localization accuracy, which explicitly captures the joint effects of link quality and anchor geometry under shared communication-localization spectrum resources. We then formulate average system throughput maximization as a mixed-integer nonlinear and non-convex optimization problem with the constraints of localization accuracy, sub-channels, UAV association, UAV deployment and signal-to-interference-plus-noise ratio. We further develop an alternating iterative optimization method to solve this complex optimization problem. Within this method, a particle swarm optimization-based method is developed to jointly optimize spectrum allocation and UAV association, and a dueling double deep Q-network-based method is further employed for UAV deployment optimization. Finally, extensive simulation results are presented to validate the efficiency of our optimization method, and also to illustrate how key parameters influence average system throughput and localization accuracy.

Joint Optimization of UAV-Mounted RIS-Assisted Mobile Edge Computing Using Deep Reinforcement Learning

Toward Low-Delay and Energy-Efficient UAV-Assisted MEC Systems Through Intelligent Resource Allocation

A Prioritized Adaptive Weighting based on Deep Deterministic Policy Gradient (PAW-DDPG) as an enhanced Deep Deterministic Policy Gradient (DDPG) algorithm to minimize both processing delay and energy consumption by jointly optimizing user scheduling, partial-task offloading, and UAV trajectory is proposed.

Reinforcement Learning-Based Decode-and-Forward UAV Relay Trajectory Optimization

Unmanned Aerial Vehicles (UAVs) are promising relay platforms due to their flexible deployment and high probability of line-of-sight (LoS) connectivity. This paper compares three deep reinforcement learning (DRL) algorithms-Proximal Policy Optimization (PPO), Soft Actor-Critic (SAC), and Recurrent PPO with LSTM memory-for joint UAV trajectory and energy optimization in UAV based relay systems. The problem formulated is a non-convex optimization problem that minimizes UAV propulsion energy while satisfying Quality of Service (QoS) and mobility constraints under realistic 3GPP channel conditions. Simulation results show that all methods achieve over 99% QoS satisfaction. SAC exhibits the fastest convergence, whereas the proposed Recurrent PPO achieves the lowest energy consumption (44.72 kJ), reducing energy usage by 5.1% compared with PPO. These results highlight the trade-off between convergence speed and energy efficiency in DRL-based UAV relay optimization.

Semantic-Oriented Image Transmission and Resource Allocation for UAV Networks

Autonomous aerial vehicles (UAVs) demonstrate significant potential for enhancing next-generation communication networks due to their flexible deployment, high adaptability, and collaborative service provision. However, the limitation of energy and communications resources hinders their widespread applications, especially for transmission of large files, e.g., high-quality images and videos. In this paper, we introduce the semantic communication technology to UAV networks for image transmission, which can extract the key semantic information and perform the maximum compression. We mathematically formulate a semantic transmission delay minimization problem, taking into account the quality standards for semantic information transmission, the limitation on network resources, and the energy consumption of each UAV. This problem is characterized as a non-convex, multi-timescale, and mixed-integer programming problem. Then, we put forth a semantic-oriented trajectory and resource allocation multi-agent reinforcement learning (SOTRA-MARL) algorithm to solve this problem, which explores the coordination of UAV trajectory, ground users’ (GUs) association strategy, and semantic information selection. The proposed algorithm allows each UAV to collaborate with others through centralized training on global information, while enabling each UAV to make distributed decisions on resource allocation during execution. Thus, this approach facilitates convergence toward a high-quality solution with fewer training iterations. Simulation results revel that our proposed algorithms significantly outperform benchmark approaches, particularly in reducing semantic information transmission delay and improving image transmission accuracy.

Semantic-Aware UAV Swarms for Low-Delay and Energy-Efficient Data Collection

Uncrewed aerial vehicle (UAV) swarms performing data collection and transmission often face heavy communication loads and high end-to-end delay, which becomes more severe when processing large volumes of raw data. Semantic communication can alleviate this bottleneck by transmitting only task-relevant information instead of full raw data, thereby reducing the communication burden. Motivated by this, we propose a joint optimization framework that integrates semantic-aware communication and computation resource allocation for UAV swarms. In the considered scenario, member UAVs extract semantic features from raw data, forward them to the cluster head UAV, and finally transmit them to the base station, where data reconstruction is performed. A long-term average total delay minimization model is formulated, and a low-complexity algorithm is developed. Specifically, the long-term problem is reformulated into a per-slot deterministic structure guided by stability, and the subproblems are solved via an alternating refinement scheme with convex-tractable closed-form updates. Simulation results show that the proposed method consistently outperforms raw-data transmission and benchmark semantic schemes across diverse settings. In particular, it reduces average total delay by 14.94% and 15.33%, and improves energy efficiency by 28.32% and 28.18% under large swarm size and high data volume, respectively.