The exploration of unmanned aerial vehicle (UAV) swarm systems represents a focal point in the research of multiagent systems, with the investigation of their fission-fusion behavior holding significant theoretical and practical value. This review systematically examines the methods for fission-fusion of UAV swarms from the perspective of multi-agent systems, encompassing the composition of UAV swarm systems and fission-fusion conditions, information interaction mechanisms, and existing fission-fusion approaches. Firstly, considering the constituent units of UAV swarms and the conditions influencing fission-fusion, this paper categorizes and introduces the UAV swarm systems. It further examines the effects and limitations of fission-fusion methods across various categories and conditions. Secondly, a comprehensive analysis of the prevalent information interaction mechanisms within UAV swarms is conducted from the perspective of information interaction structures. The advantages and limitations of various mechanisms in the context of fission-fusion behaviors are summarized and synthesized. Thirdly, this paper consolidates the existing implementation research findings related to the fission-fusion behavior of UAV swarms, identifies unresolved issues in fission-fusion research, and discusses potential solutions. Finally, the paper concludes with a comprehensive summary and systematically outlines future research opportunities.

This paper investigates a measurement-driven rotary-wing unmanned aerial vehicle (UAV)-enabled wireless communication system serving multiple ground users. Unlike conventional studies relying on simplified theoretical channel models, we derive the air-to-ground (A2G) path loss and shadow-fading models from software-defined radio (SDR) channel measurements at 3.3 GHz. To maximize the minimum average user rate, we formulate a joint optimization of three-dimensional (3D) trajectory, velocity, transmit power, and bandwidth, subject to quality-of-service (QoS) and UAV kinematic constraints. Owing to the strong coupling and non-convexity, we reformulate it as a constrained Markov decision process (CMDP) with a continuous-action space and solve it via proximal policy optimization (PPO), where a constraint-preserving action parameterization enforces power, bandwidth, and motion constraints by construction. Simulations over real digital-elevation-model terrain show that the proposed approach attains 95.7% of a search-based rate reference at $K=3$ and surpasses an offline AO/SCA benchmark in max–min rate at all user densities, with 100% QoS satisfaction at $K=3$, near-zero constraint violations, and $0.142$–$0.181$ ms online inference per decision.

Air-to-ground (A2G) channel modeling supports low-altitude wireless networks for unmanned aerial vehicle (UAV) services in 5G/6G systems. Existing data-driven models are often benchmarked against weak empirical baselines, decoupled from the underlying measurement campaigns, and prone to overfitting on small UAV datasets. This paper presents PACE-A2G, a calibrated deep-tree ensemble on physics-aware features, together with a stricter evaluation methodology. A quadrotor-mounted software-defined-radio measurement system produces five datasets covering 433, 740, 1500, and 3300 MHz in overwater, suburban, and mountainous environments. The deep tabular backbone combines periodic numerical embeddings, feature-group attention, and a parameter-efficient ensembling head; propagation physics enters through the features rather than through the architecture. At inference, the deep estimate is blended with a tree ensemble, and a split-conformal quantile head provides calibrated prediction intervals. Evaluation under random, chronological, and spatial-block protocols shows that random partitions overestimate accuracy: the chronological-split root-mean-square error (RMSE) of the strongest tree-ensemble baseline is 1.9 to 8 times its random-split value. In distribution, the hybrid predictor is statistically on par with the strongest tree ensembles. Across nine cross-scenario and cross-frequency protocols, it transfers more robustly than the tree ensemble on balance, with a mean paired RMSE reduction of 0.34 dB, although the fitted log-distance baseline remains the strongest single predictor under shift. Split-conformal calibration reaches near-nominal 90% coverage in distribution but collapses under domain shift; a feature-space out-of-distribution gate detects the shift on every source–target pair, and ten labelled records of the new domain restore near-nominal coverage.