Power management and performance optimization of underwater wireless sensor networks based on MARL

Area of Science:

• Marine Technology
• Computer Science
• Network Engineering

Background:

• Underwater wireless sensor networks (UWSNs) face performance degradation due to complex channels and limited resources.
• Existing methods struggle to balance energy consumption, transmission reliability, and network longevity.

Purpose of the Study:

• To propose an intelligent algorithm for optimizing UWSN performance.
• To reduce energy consumption, enhance reliability, and extend network lifetime.

Main Methods:

• Formalized UWSN optimization as a partially observable Markov decision process.
• Applied multi-agent reinforcement learning with a team reward function (fair reuse rewards, survival time penalties).
• Developed a distributed intelligent power management scheme for nodes.

Main Results:

• Achieved network capacity of 245.68 kb and fairness reuse index of 1.85 in heterogeneous scenarios.
• Maintained low communication latency (6.18 time slots) with 5% node failures.
• Demonstrated robustness in dynamic environments, sustaining network capacity over 32,045 kb and energy efficiency of 0.4 kb/J.

Conclusions:

• The proposed multi-agent reinforcement learning algorithm significantly improves UWSN robustness and performance.
• This approach offers effective communication support for ocean monitoring applications.
• Intelligent power management enhances network lifetime and reliability in challenging underwater conditions.