With the advancements in sensing technologies, maritime sensing has become indispensable in various domains, including logistics, weather forecasting, and marine ranching. However, transmitting large volumes of sensing data faces many challenges in the maritime environment. First, the transmissions purely depend on satellite links often costly and suffer from long propagation latency. On the other hand, traditional unicast transmission results in data duplication, wasting valuable marine communication resources. With the increasing density of sensing devices, the communication distance between maritime sensors has become closer, enabling the deployment of maritime opportunistic networks consisting of device-to-device links. Rather than using unicast transmission over satellite links, employing multicast with opportunistic routing enables simultaneous data transmission to multiple destinations and saves communication resources. Even though the multicast method can avoid redundancy, conducting multicast without considering the maritime characteristics (i.e., the dynamics and the distribution of sensors) may lead to inefficient data delivery. Through real-world experiments, we observe that devices located on the edges of the network have a relatively low receiving rate compared with internal ones and tend to be the bottleneck of the overall multicast progress. Based on this observation, we propose BOOM, a bottleneck-aware opportunistic multicast strategy aiming at reducing multicast latency, taking into account the influence of the bottleneck node and broadcasting rate. Prominently, within maritime scenarios challenged by extreme conditions, such as storms, typhoons, and tsunamis, BOOM’s emphasis encompasses the adaptability of multicast strategies, which necessitates dynamic adjustments in response to equipment failures and shifts in network topology. Through mathematical analysis, we prove the formation of opportunistic multicast is an NP-hard problem and further design a heuristic algorithm based on the convex-hull method to reduce the computational cost in strategy generation. We compare BOOM with four other algorithms using real-world maritime vessel trajectories in various scenarios. The simulation result illustrates that the BOOM achieves a significant reduction in transmission latency, which reduces 36% when sensors are sparsely located in water areas, and the reduction could reach up to 59% when sensors are more dense. Furthermore, in extreme environmental testing conditions, BOOM continues to outperform other algorithms in terms of completion time, with performance improvements of up to 39% and 49% in sparse and dense topology environments, respectively.

Satellite battery status prediction is crucial for ensuring the healthy operation of future satellite constellations. However, traditional telemetry-based methods, where satellite battery status is transmitted in real time to ground stations for processing, consume significant satellite bandwidth and introduce response delays. Advances in onboard computing and federated learning (FL) enable local model training and centralized parameter aggregation, reducing transmission overhead while leveraging distributed satellite data. Nevertheless, the unique orbital motion of satellites presents challenges for FL, primarily due to battery status heterogeneity arising from varying sunlight exposure. Limited onboard energy further necessitates balancing model performance with battery efficiency during local training. To tackle these issues, we propose ObsBattery—a position-aware FL framework that clusters satellites based on their orbital positions to improve model accuracy. ObsBattery employs a Dueling Deep Q-Network to dynamically determine satellite clustering and adapt local training rounds according to power availability, thereby reducing energy consumption during low-power phases. Evaluations on a real-world satellite battery dataset show that ObsBattery significantly improves both prediction accuracy and energy efficiency. Compared to a standard clustered FL approach, it reduces model MAE by 16% and energy consumption ratio by 6% under experimental conditions.

In resource-constrained vehicular environments, offloading tasks to edge servers simultaneously from multiple client vehicles leads to severe resource competition, causing a cycle of increased latency. Designing a task offloading strategy that balances the latency of new offloaded tasks and those already being executed on edge servers is a crucial challenge. Additionally, centralized task offloading strategies based on global information require the design of complex communication mechanisms to collect task and computational workload information in real time, which is not desirable in vehicular environments due to the dynamic changes in the locations of client vehicles and the varying task offloading demands across time and space. To address these issues in multi-client vehicular task offloading environments, we propose Moscato, a blockchain-based distributed task offloading framework. In Moscato, client vehicles function as blockchain consensus nodes, utilizing existing consensus mechanisms for asynchronous, non-real-time global information sharing. To optimize task offloading decisions under diverse task profiles and dynamic traffic conditions, we integrate Federated Learning (FL) with Deep Q-Network (DQN), enabling intelligent, decentralized decision-making. Real-world datasets on edge server workloads and task latencies were collected to conduct simulation-based evaluations. Through comparison with state-of-the-art methods, we demonstrate that Moscato can design a better balance between the execution of newly offloaded tasks and ongoing ones, effectively alleviating resource competition under multi-client scenarios.

The rapid development of UnmannedAerial Vehicle (UAV) swarm technologies has significantly expanded their applications in fields like road rescue and agricultural monitoring, where multi-modal data streams from sensors such as infrared, radar, and visible light are critical. However, processing such data in real-time is computationally intensive, particularly with UAVs' limited onboard resources. Missions involving multi-modal data are typically divided into sub-tasks, which can be processed either locally or offloaded to edge servers. While offloading sub-tasks helps alleviate onboard computational load, it introduces transmission delays and potential server overloads. Moreover, the inter-dependency of sub-tasks creates a synchronization bottleneck, where the slowest sub-task determines the mission completion time. Early-completed sub-tasks also occupy limited UAV onboard buffer resources. To address these challenges, we propose LABOR, a Proximal Policy Optimization (PPO)-based mission offloading strategy. LABOR dynamically determines whether each sub-task should be processed locally or offloaded, optimizing the balance between latency, resource availability, and computational requirements. Through extensive simulations in various scenarios, LABOR is shown to significantly outperform other offloading strategies, achieving up to 64.0% and 80.1% reductions in latency and buffer occupancy, respectively. This work provides a promising solution for efficient mission offloading in resource-constrained UAV swarm networks, enhancing both real-time performance and resource utilization.

The transition from physical network functions (PNFs) to virtual network functions (VNFs) and cloud-native network functions (CNFs) represents a dynamic advancement in virtualization technologies, addressing limitations in flexibility, prolonged deployment times, and scalability. CNFs encapsulate functionalities in lightweight, containerized units, prioritizing cloud-native principles of agility and resilience. Effective management of CNFs is paramount for modern network infrastructures, and Open Source Management and Orchestration (OSM) frameworks aim to tackle these challenges. However, certain constraints persist, particularly in large-scale applications with complex startup sequencing. These challenges are further complicated by the modular nature of CNFs and the dynamic resource allocation in cloud-native environments. To address these issues, we introduce Homcs, a hierarchical orchestration mechanism for CNF services in 5G Core Networks. Homcs leverages Helm for orchestrated deployment and implements a RESTful Information Exposure Interface within Kubernetes. Real-world experiments have demonstrated that Homcs enhances CNF orchestration by optimizing deployment, scalability, and resource allocation within OSM environments.