Minimizing Maximum Latency of Task Offloading for Multi-UAV-assisted Maritime Search and Rescue

Shuang Qi, Bin Lin, Yiqin Deng, Xianhao Chen, Yuguang Fang

Abstract: In Maritime Search and Rescue (MSAR) operations, timely information is critical for saving lives during the "Golden Hour." Traditional communication infrastructures often fail to provide the high bandwidth required for real-time video surveillance in remote ocean areas. This paper presents a multi-UAV-assisted mobile edge computing (MEC) framework designed to minimize the maximum latency of task offloading. By deploying a fleet of surveillance UAVs (S-UAVs) and a high-capacity relay UAV (R-UAV), the system optimizes target association, UAV deployment, and computation offloading. Simulation results demonstrate that this coordinated approach significantly reduces processing delays and improves synchronization compared to traditional methods, offering a robust solution for time-sensitive maritime rescue missions.

Introduction

Maritime Search and Rescue (MSAR) is a race against time. When a vessel is distressed, the vastness of the sea and unpredictable weather conditions make locating survivors a daunting task. While Unmanned Aerial Vehicles (UAVs) have emerged as powerful tools for rapid aerial surveillance, their effectiveness is often limited by constrained onboard battery life and processing power. Furthermore, transmitting high-definition video back to shore via satellite or long-range radio introduces significant latency, which can delay critical decision-making.

The integrated maritime search and rescue communication network, featuring satellites, UAVs, and shore stations.

To address these challenges, researchers have proposed integrating Mobile Edge Computing (MEC) into UAV networks. By offloading data-intensive tasks—such as real-time image recognition and survivor tracking—to nearby aerial servers, the system can bypass the bottlenecks of traditional communication links.

System Architecture and Methodology

The proposed framework utilizes a two-tier UAV architecture to manage the disaster area effectively:

  1. Surveillance UAVs (S-UAVs): These are lightweight drones equipped with cameras to monitor specific rescue targets. They perform initial data collection and can process some video chunks locally.
  2. Relay UAV (R-UAV): Positioned at a higher altitude, the R-UAV serves as a powerful edge server and a communication bridge between the S-UAVs and the coastal Rescue Coordination Center (RCC).
The two-tier UAV architecture showing S-UAVs monitoring targets and offloading tasks to a central Relay UAV.

The core of this research is the joint optimization of three critical variables: target association, UAV deployment, and offloading decisions. This is formulated as a Mixed-Integer Nonlinear Programming (MINLP) problem aimed at minimizing the maximum latency among all S-UAVs. This "min-max" approach ensures fairness and synchronization, preventing any single drone from becoming a bottleneck that delays the overall rescue picture.

To solve this complex problem, the authors decomposed it into three sub-problems:

  • Offloading Optimization: Determining the optimal ratio of local vs. remote processing for each drone.
  • R-UAV Deployment: Using Successive Convex Approximation (SCA) to find the 3D coordinates for the R-UAV that maximize the signal-to-interference-plus-noise ratio (SINR) for the entire fleet.
  • Target Association: Employing a Branch and Bound (BnB) algorithm to assign S-UAVs to targets, ensuring full coverage while balancing the computational load.

Key Insights and Performance

A pivotal finding of the study is the importance of the R-UAV's computational capacity. As the R-UAV is equipped to handle more simultaneous tasks, the system's overall latency drops dramatically. Additionally, the S-UAVs dynamically adjust their Field of View (FOV) by changing altitude; they fly lower for high-resolution tracking of clustered targets and higher to maintain a broader view of dispersed survivors.

Numerical simulations modeling a 1 km² area with 20 targets showed that the proposed algorithm consistently outperforms benchmark schemes. Specifically, it maintains lower latency even as the volume of video data increases, proving its scalability for high-stakes environments. While the energy consumption for transmission is higher than local-only processing, it remains well within the operational limits of modern UAV batteries.

Conclusion

By synchronizing the "flying edge," this multi-UAV framework provides a faster, more reliable way to conduct maritime rescues. Future research will look into multi-hop communication for larger disaster zones and further optimizing energy efficiency to prolong mission endurance. This work marks a significant step toward a fully automated, high-speed aerial rescue infrastructure.

Summary updated

BibTeX
@article{qi2024minimizing,
  author = {Shuang Qi and Bin Lin and Yiqin Deng and Xianhao Chen and Yuguang Fang},
  title = {{Minimizing Maximum Latency of Task Offloading for Multi-UAV-assisted Maritime Search and Rescue}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2024},
  doi = {10.1109/tvt.2024.3384570},
}
Figure