Research on Computation Offloading and Resource Allocation Strategy Based on MADDPG for Integrated Space-Air-Marine...

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Maritime Internet of Things (M-IoT) networks face challenges in computation offloading and resource allocation due to complex, dynamic environments.
• Traditional optimization methods struggle with non-convexity and combinatorial complexity in these integrated space-air-sea networks.

Purpose of the Study:

• To develop an efficient optimization algorithm for computation offloading and resource allocation in integrated space-air-sea networks.
• To minimize total system costs, balancing energy consumption and latency for M-IoT devices.

Main Methods:

• A multi-agent deep deterministic policy gradient (MADDPG)-based optimization algorithm was proposed.
• The system was modeled as a partially observable Markov decision process (POMDP), integrating association, power, computing resource allocation, and task distribution.
• A centralized training with decentralized execution framework was employed, with M-IoT devices and UAVs acting as intelligent agents.

Main Results:

• The proposed MADDPG-based algorithm demonstrated rapid convergence.
• Numerical simulations showed significant outperformance compared to baseline methods.
• The algorithm achieved a reduction in total system cost by 15-60%.

Conclusions:

• The MADDPG-based approach is effective for optimizing computation offloading and resource allocation in integrated space-air-sea networks.
• The algorithm successfully balances energy consumption and latency through partial task offloading.
• This method offers a robust solution for enhancing M-IoT device performance in challenging maritime environments.