SA-TD3: UAV-assisted task offloading and optimized resource allocation for maritime edge computing.

Area of Science:

• Marine Internet of Things
• UAV-assisted computation offloading
• Resource optimization

Background:

• Existing research on UAV-assisted task offloading in Marine IoT primarily focuses on UAV trajectory and resource allocation.
• Dynamic demands and optimization potential of maritime surface nodes are often overlooked.
• Static resource allocation and single optimization methods limit adaptability in dynamic ocean environments.

Purpose of the Study:

• To propose a hybrid decision-making framework (SA-TD3) for UAV-assisted computation offloading and resource optimization from the perspective of maritime surface nodes.
• To address dynamic task demands and enhance global search capability and policy robustness.
• To improve training efficiency by capturing inter-node correlations.

Main Methods:

• Developed a UAV-assisted computation offloading and resource optimization mechanism tailored for maritime surface nodes.
• Enhanced the TD3 algorithm by integrating simulated annealing and an environment-aware dual-channel advantage function.
• Implemented a graph neural network-based dynamic prioritized replay mechanism for inter-node correlation capture.

Main Results:

• SA-TD3 reduced average latency by 19.7%.
• Overall performance improved by 13.2% across diverse ocean environments.
• Demonstrated effective reduction in computational load and communication latency for surface nodes, alongside enhanced energy efficiency.

Conclusions:

• SA-TD3 offers a robust and adaptable solution for UAV-assisted task offloading in Marine IoT.
• The framework effectively addresses dynamic task demands and optimizes resource allocation for maritime surface nodes.
• The proposed methods significantly improve system performance, latency, and energy efficiency in complex ocean environments.