Multi-Copy Relay Node Selection Strategy Based on Reinforcement Learning.

Area of Science:

• Computer Science
• Network Engineering
• Artificial Intelligence

Background:

• Delay Tolerant Networks (DTNs) face challenges with path establishment and significant message delays.
• Efficient relay node selection is crucial for managing network overhead, reducing latency, and enhancing delivery rates in DTNs.

Purpose of the Study:

• To develop a multi-copy relay node selection algorithm for DTNs.
• To improve network performance by reducing overhead and delays while increasing delivery rates.

Main Methods:

• The proposed algorithm, Q-lambda reinforcement learning with community division (QLCR), utilizes community division principles to identify core nodes based on degree.
• Relay node selection incorporates node degree, interests, and structural similarity.
• Q-lambda reinforcement learning enables nodes to learn network-wide, assigning rewards to optimize relay node selection through iterative processes.

Main Results:

• The QLCR algorithm demonstrated a high message delivery rate in experimental evaluations.
• The algorithm effectively maintained low network overhead and reduced message delays.

Conclusions:

• The QLCR algorithm offers a robust solution for relay node selection in Delay Tolerant Networks.
• This approach significantly enhances DTN performance metrics, including delivery rate, overhead, and delay.