Self-Organizing Wireless Sensor Networks Solving the Coverage Problem: Game-Theoretic Learning Automata and Cellular...

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Wireless Sensor Networks (WSNs) face challenges in distributed coverage optimization.
• Existing solutions often require central coordination, which is not ideal for WSNs.

Purpose of the Study:

• To develop self-organizing algorithms for distributed coverage in WSNs.
• To apply a game-theoretical framework using agent-based reinforcement learning.

Main Methods:

• A game-theoretical framework based on the Spatial Prisoner's Dilemma game was adapted.
• Multi-agent systems were modeled using graph theory, with nodes as agents.
• Two agent types were used: Learning Automata (LA) and Cellular Automata (CA).
• A novel (ϵ,h)-learning automata agent model was developed and compared to adaptive CA agents.

Main Results:

• Agents reaching Nash equilibria in iterated games led to global optimization of coverage and sensor usage.
• The (ϵ,h)-learning automata model significantly outperformed the cellular automata model.
• Distributed coverage was achieved without agents knowing the global criterion or needing a central coordinator.

Conclusions:

• Game-theoretical multi-agent systems offer an effective approach for distributed WSN coverage.
• The proposed (ϵ,h)-learning automata model provides a superior solution compared to CA for this problem.