Joint Optimization of Time Slot and Power Allocation in Underwater Acoustic Communication Networks

Area of Science:

• Underwater Acoustic Communication Networks (UACNs)
• Reinforcement Learning (RL)
• Optimization Algorithms

Background:

• Underwater acoustic communication networks face challenges in efficient time slot and power allocation.
• Existing methods like TDMA and Slotted ALOHA have limitations in maximizing network capacity and managing collisions.

Purpose of the Study:

• To propose a joint optimization algorithm for time slot and power allocation in UACNs.
• To maximize the total capacity of successful transmissions in underwater acoustic communication.

Main Methods:

• Developed a joint optimization algorithm using reinforcement learning.
• Formulated two sub-objectives for time-slot scheduling (using Deep Q-Network) and power allocation (using Multi-Agent Deep Deterministic Policy Gradient).
• Implemented centralized training with distributed execution for power allocation.

Main Results:

• The proposed algorithm significantly improves the number of successfully transmitted links.
• Demonstrated enhanced channel capacity compared to traditional algorithms like TDMA and Slotted ALOHA.
• Effectively managed energy limitations while maximizing transmission capacity.

Conclusions:

• The joint optimization algorithm offers superior performance for UACNs.
• Reinforcement learning, specifically DQN and MADDPG, provides an effective framework for UACN resource allocation.
• The approach successfully balances throughput and energy efficiency in underwater acoustic communication.