[Paper] Double-Edge-Assisted Computation Offloading and Resource Allocation for Space-Air-Marine Integrated Networks

Source: arXiv - 2512.03487v1

Overview

The paper introduces a double‑edge‑assisted framework that lets maritime autonomous surface ships (MASS) offload part of their compute‑intensive tasks to two different edge nodes simultaneously: a fleet of UAVs hovering above the sea and a low‑Earth‑orbit (LEO) satellite. By jointly deciding where to offload, how much to offload, and how to allocate the limited computing resources on the UAVs and satellite, the authors aim to cut the overall energy consumption while respecting strict latency requirements.

Key Contributions

• Dual‑edge architecture for Space‑Air‑Marine Integrated Networks (SAMINs) that leverages both aerial (UAV) and space (LEO) edge servers.
• Joint optimization model that simultaneously selects offloading mode, offloading volume, and computing resource allocation to minimize total energy under latency constraints.
• Algorithmic solution based on Alternating Optimization (AO) and a layered decomposition that yields near‑optimal solutions with tractable complexity.
• Comprehensive simulation study comparing the proposed scheme against several baselines, demonstrating up to ~30 % energy savings and improved latency compliance.

Methodology

1. System Model – The authors model a set of MASSs that generate computation tasks with known data size and deadline. Each task can be split: a fraction runs locally on the ship, another fraction is sent to a UAV, and the remainder to the LEO satellite. Communication links (ship‑UAV, ship‑satellite) are characterized by bandwidth, channel gain, and propagation delay.

2. Energy‑Latency Formulation – The total energy consists of three parts:

   • (i) local CPU energy on the ship,
   • (ii) transmission energy for the offloaded bits, and
   • (iii) processing energy on the edge nodes.
     The latency constraint ensures that the sum of transmission time and edge‑processing time for each offloaded portion does not exceed the task deadline.

3. Optimization Problem – The goal is to minimize total energy by choosing:

   • Offloading mode (binary decision for each edge node),
   • Offloading volume (how many bits go to each node), and
   • Computing resource allocation (CPU cycles per second allocated to each ship on the UAV and satellite).
     The problem is non‑convex because of coupled variables and integer decisions.

4. Solution Approach

   • Alternating Optimization (AO): The problem is split into two sub‑problems that are solved iteratively: (a) offloading decisions & volumes (continuous) and (b) CPU‑resource allocation (convex).
   • Layered Decomposition: Within each AO step, the authors further decompose the sub‑problem into a master problem (binary mode selection) and a slave problem (continuous volume allocation) that can be solved via standard convex solvers.
   • Convergence: The AO loop is proven to converge to a stationary point, and the layered approach guarantees that each sub‑problem reaches its local optimum.

Results & Findings

• Energy Savings: Compared with three benchmarks (local‑only, single‑edge UAV, single‑edge satellite), the double‑edge scheme reduces total energy consumption by 23 %–31 % across a range of task sizes and deadlines.
• Latency Compliance: The proposed method meets the latency constraints for >95 % of tasks, whereas single‑edge solutions violate deadlines in up to 18 % of cases under heavy load.
• Scalability: Simulation with up to 20 ships shows that the algorithm’s runtime grows linearly, making it practical for real‑time orchestration.
• Resource Utilization: The UAVs handle latency‑sensitive, smaller workloads, while the LEO satellite absorbs larger, less time‑critical portions, achieving a balanced load distribution.

Practical Implications

• Maritime IoT & Autonomous Shipping: Operators can extend the computational reach of autonomous vessels without over‑provisioning onboard hardware, leading to lighter, cheaper ships.
• Edge‑Enabled 5G/6G Networks: The dual‑edge concept fits naturally into emerging 5G‑Advanced and 6G architectures that envision integrated terrestrial, aerial, and satellite layers.
• Energy‑Constrained Deployments: For battery‑powered ships or UAVs, the joint offloading strategy can dramatically prolong mission endurance.
• API‑Level Integration: The optimization framework can be encapsulated as a service (e.g., “edge‑offload orchestrator”) that takes task descriptors and returns a split‑offload plan, enabling developers to plug it into existing maritime control software.

Limitations & Future Work

• Channel Uncertainty: The model assumes perfect knowledge of link quality; real‑world fading and weather effects (especially for UAV‑to‑ship links) could degrade performance.
• Static UAV/LEO Placement: The study treats UAV positions and satellite beams as fixed; dynamic repositioning or beamforming could further improve latency and energy metrics.
• Security & Trust: Offloading to aerial and space nodes raises authentication and data‑privacy concerns that are not addressed.
• Future Directions: Extending the framework to incorporate stochastic channel models, adaptive UAV trajectory planning, and secure multi‑party computation would bring the solution closer to deployment in operational SAMINs.

Authors

• Zhen Wang
• Bin Lin
• Qiang
• Ye

Paper Information

• arXiv ID: 2512.03487v1
• Categories: cs.DC, cs.IT
• Published: December 3, 2025
• PDF: Download PDF