Using neural networks to solve the multicast routing problem in packet radio networks

Area of Science:

• Computer Science
• Network Engineering
• Artificial Intelligence

Background:

• Packet radio networks prioritize efficient information transfer, minimizing bandwidth and end-to-end delay.
• Existing research primarily addresses single-source, single-destination routing, with limited focus on multicasting.
• Multicasting, sending a packet to multiple destinations, presents unique routing challenges, especially when considering packet delay.

Purpose of the Study:

• To develop a practical technique for computing optimal or near-optimal multicasting solutions in packet radio networks.
• To address the complexities introduced by including packet delay in network topology for multicasting.
• To provide an efficient routing mechanism for scenarios requiring data dissemination to multiple network nodes.

Main Methods:

• The study employs a Hopfield neural network model to solve the multicasting routing problem.
• The technique is designed to handle both scenarios: with and without considering packet delay at each node.
• Computational experiments were conducted to evaluate the performance of the proposed method.

Main Results:

• The Hopfield neural network approach yields near-optimal solutions for the multicasting problem.
• The method effectively minimizes both the number of transmissions and end-to-end delay.
• The technique requires a minimal amount of computational time (CPU time), demonstrating practical efficiency.

Conclusions:

• The proposed Hopfield neural network-based technique offers an effective solution for multicasting in packet radio networks.
• This approach provides a practical and computationally efficient way to achieve optimal or near-optimal multicasting.
• The method advances network routing strategies by addressing the complex challenge of efficient multi-destination packet delivery.