Reinforcement Learning-Based Data Forwarding in Underwater Wireless Sensor Networks with Passive Mobility.

Area of Science:

• Computer Science
• Engineering
• Marine Technology

Background:

• Underwater wireless sensor networks (UWSNs) face significant challenges in data forwarding due to harsh communication environments.
• Key issues include ensuring data timeliness and conserving limited node energy, which impacts network lifetime.
• The passive mobility of underwater sensors introduces dynamic topology changes, further complicating network management.

Purpose of the Study:

• To formulate the data forwarding problem in UWSNs by considering timeliness, energy consumption, and passive node mobility.
• To propose an effective solution using reinforcement learning for UWSN data forwarding.
• To evaluate the performance of the proposed method against benchmark protocols.

Main Methods:

• Formulation of the data forwarding problem, integrating timeliness, energy efficiency, and a passive mobility model.
• Development of a reinforcement learning-based algorithm to address the formulated problem.
• Performance evaluation through extensive simulations to validate the proposed method.

Main Results:

• The proposed reinforcement learning method effectively addresses the challenges of data forwarding in UWSNs.
• Simulation results demonstrate superior performance compared to benchmark protocols in terms of timeliness and energy efficiency.
• The method achieved an 83.35% increase in the value of information and up to 75.21% energy savings compared to the QELAR protocol.

Conclusions:

• The reinforcement learning-based approach is a valid and effective solution for data forwarding in UWSNs.
• The proposed method significantly improves both data delivery timeliness and network energy efficiency.
• This work contributes to advancing the capabilities of underwater sensor networks for various applications.