Muhammad Yeasir Arafat

Multi-tier aerial communication networks (MACNs), integrating satellites, high-altitude platforms, and unmanned aerial vehicles, are emerging as a cornerstone of next-generation global connectivity. Their promise of resilient and ubiquitous coverage, however, is hindered by highly dynamic topologies, severe energy and computational constraints, environment-sensitive channels, diverse quality-of-service requirements, and limited real-world validation. Artificial intelligence (AI) has increasingly been explored as a flexible framework to address these challenges, enabling adaptive routing, distributed computing and task offloading, handover management, intelligent resource allocation, and large-scale network optimization. This survey provides a comprehensive and structured review of methods for MACNs, with particular emphasis on AI-driven solutions and their relationship to classical and hybrid alternatives. We critically evaluate representative approaches in terms of scalability, efficiency, data demands, and practical deployability, and identify emerging trends such as graph neural networks with reinforcement learning for dynamic routing, predictive learning for mobility management, and federated learning for distributed computation. Persistent challenges remain in lightweight edge intelligence, real-world testbeds, reproducible benchmarking, and simulation-to-deployment transfer. To address these issues, we offer a research roadmap emphasizing compressible and interpretable models, standardized benchmarks, realistic validation, and hybrid designs that balance adaptability with computational and energy overhead. Finally, we identify open challenges and future research directions, offering insights into the design of AI-driven MACNs that are efficient, scalable, and adaptive to evolving network and service demands.

In mission-critical environments, designing efficient routing protocols for unmanned aerial vehicles (UAVs) is challenging due to limited energy resources, high-speed mobility, and highly dynamic link conditions. This paper proposes a hierarchical routing framework, termed hierarchically coordinated proximal multi-hop routing (HCPMR) scheme, which integrates multi-agent proximal policy optimization (MAPPO) with graph neural networks (GNNs) to enable scalable, stable, and energy-efficient UAV communication. HCPMR organizes UAVs into energy- and stability-aware clusters, where GNN- assisted intra-cluster coordination is performed by cluster leaders, who also manage inter-cluster routing through periodically synchronized exchanges under a centralized training and decentralized execution (CTDE) paradigm. GNN-based embeddings allow UAVs to capture spatial dependencies within their local neighbourhoods, enabling adaptive routing with reduced control overhead, enhanced reliability, and improved energy efficiency. The proposed scheme incorporates a multi-objective reward function to jointly optimize energy consumption, fairness, delivery reliability, and terrain adaptability. Simulation results using NS-3/ns3-gym across urban, desert, and mountainous scenarios demonstrate that HCPMR increases packet delivery ratio by 6.5% compared to coordinated Q-learning-based multi-hop routing (CQMR) and by 16% compared to improved Q-learning–based multi-hop routing (IQMR). Moreover, HCPMR achieves approximately 13% higher residual energy than CQMR and 58% higher than IQMR, while reducing control overhead by 20% and 38.5%, respectively. These results establish HCPMR as an adaptive and efficient routing framework for dynamic FANETs in mission-critical operations, effectively enhancing routing decisions through GNN-driven embeddings and MAPPO-based coordination.

In air–ground collaborative networks, low-altitude unmanned aerial vehicle (UAV) swarms and high-altitude platform cooperate to ensure reliable and efficient data collection for ground users (GUs) in dynamic and mission-critical environments. Maintaining data freshness, quantified by the age of information (AoI), while minimizing energy consumption is highly challenging due to UAV mobility, heterogeneous agents, and limited resources. This paper formulates a joint optimization problem that simultaneously minimizes the long-term average AoI and energy consumption of UAVs and GUs under realistic constraints, including quality of service (QoS), queue size, and resource limitations. To tackle this complex mixed discrete–continuous problem, we propose a swarming behavior-integrated multi-agent multi-expert soft actor-critic (MA-MESAC) framework that synergizes swarm intelligence with multi-agent deep reinforcement learning. The actor network employs self-attention to model intra-agent dependencies, while the critic network uses cross-attention to capture inter-agent spatial–temporal correlations. We design a composite actor loss function by integrating the SAC-based policy objective with regularization terms that enforce fair resource allocation and physics-informed swarming behavior, ensuring stable and efficient learning. Moreover, we design a multi-objective reward function that explicitly accounts for stringent mobility, QoS, and resource constraints. Simulation results demonstrate that the proposed framework significantly enhances fairness-aware data freshness, energy efficiency, and convergence compared with existing baselines.

Reliable video streaming is a critical requirement for emerging unmanned aerial vehicle (UAV) applications such as surveillance, disaster response, and industrial inspection. Achieving this requirement is challenging under dynamic wireless channel conditions and varying propagation environments. This study evaluates Luby transform (LT)‐fountain‐coded UAV video streaming over 5.9 GHz and 30 GHz links in outdoor urban macro (UMa) and indoor factory (InF) environments. Standardized 3GPP channel models are applied under both line‐of‐sight (LoS) and mixed LoS/non‐line‐of‐sight (NLoS) conditions. The evaluation distinguishes packet‐level wireless reliability from video‐streaming performance. Packet‐level reliability is characterized using signal‐to‐noise ratio and packet‐erasure behaviour, while video‐streaming performance is evaluated using frame recovery success rate, LT‐coded symbol demand, effective coding rate, delivered application‐layer throughput, and coded‐frame delivery latency. LT coding is assessed in terms of its ability to recover video frames under erasure conditions, its required redundancy overhead, and its impact on decoding success and effective rate. Simulation results show that 5.9 GHz links maintain more robust connectivity over longer distances, achieving lower packet‐erasure rates and higher frame recovery success rates. The 30 GHz configuration offers higher nominal bandwidth but is more sensitive to path loss, blockage, and LoS/NLoS transitions, particularly in InF and NLoS scenarios. These findings highlight the trade‐offs between wireless robustness and video‐level performance across frequency bands and environments. The results further show that LT fountain coding improves frame‐recovery reliability while maintaining sub‐second coded‐frame delivery latency under the adopted contention‐free scheduled‐link abstraction.

In recent years, unmanned aerial vehicles (UAVs), commonly referred to as drones, have witnessed rapid growth in adoption across both civilian and military domains, owing to their computational efficiency, flexibility, and adaptive control capabilities. With the growing global demand for real-time surveillance, UAV-based surveillance systems have emerged as a promising solution to enhance situational awareness, public safety, and operational efficiency across diverse environments. This systematic review evaluates recent advancements, applications, and trends in UAV-based surveillance technologies. The paper outlines the literature selection methodology, examines the current market landscape, highlights key technological innovations, sensor technologies, deployment strategies, computational frameworks, and bridges the gap between academic research and industry practice. The review highlights innovations in lightweight and full-scale artificial intelligence models, edge computing solutions, and autonomous navigation, while identifying persistent challenges related to energy efficiency, latency, and privacy. A comparative analysis elucidates performance trade-offs between commercial and research prototypes. Finally, this paper explores emerging research directions and offers insights to guide future developments in the design of intelligent, reliable, and scalable UAV-based surveillance systems.

Wireless Body Area Networks (WBANs) are recognized as innovative technology for personal health monitoring. In WBANs, physiological sensor data must be transmitted to the local processing unit (LPU) with minimal age of information (AoI) based on sensor data priority. However, the limited resources of sensors, such as energy, computational capacity, caching and bandwidth, make AoI minimization challenging. Additionally, dynamic radio links caused by body movement and interference further complicate the task. This study aims to minimize the weighted cost of time average AoI and energy consumption by adaptively controlling the transmit power based on real-time distance variations between the sensor and LPU, while allocating priority-aware bandwidth under quality of service and resource constraints in continuous decision space. To solve the non-linear problem, a particle filter-assisted Lagrange relaxation with Karush–Kuhn–Tucker conditions (PF-LKKT) framework is proposed. A recursive particle filter based on received signal strength and motion-sensor data is employed for accurate distance estimation. Then, according to the distance variation adaptively transmit power and data priority-aware bandwidth is jointly allocated. Simulation results demonstrate the superiority of the proposed framework over existing techniques.