Optimizing coverage in wireless sensor networks using deep reinforcement learning with graph neural networks

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Achieving optimal coverage in dynamic Wireless Sensor Networks (WSNs) is a persistent challenge.
• Traditional optimization methods lack real-time self-learning and require frequent retraining for changing conditions.

Purpose of the Study:

• To introduce a novel hybrid model integrating Deep Reinforcement Learning (DRL) and Graph Neural Networks (GNN) for enhanced WSN coverage.
• To improve real-time adaptability and operational efficiency in dynamic WSN environments.

Main Methods:

• Integration of Deep Reinforcement Learning (DRL) for adaptive decision-making and real-time sensor node adjustments.
• Utilization of Graph Neural Networks (GNN) to capture spatial dependencies and optimize coverage.
• Extensive simulations to evaluate the performance of the proposed DRL-GNN model.

Main Results:

• The DRL-GNN model achieved a coverage ratio of up to 96.4%.
• Demonstrated energy efficiency of 95.8% in simulated environments.
• Minimized network overlap to 5.2%, outperforming traditional optimization techniques.

Conclusions:

• The proposed DRL-GNN model effectively enhances WSN coverage and operational efficiency.
• The hybrid approach maintains high energy efficiency while minimizing redundancy in dynamic settings.
• This research validates the effectiveness of combining DRL and GNN for advanced WSN optimization.