AI-assisted intent-based traffic grooming in a dynamically shared 5g optical fronthaul network

Area of Science:

• Telecommunications Engineering
• Network Science
• Artificial Intelligence

Background:

• Optical fronthaul networks struggle with increasing bandwidth demands and dynamic traffic patterns from cell sites.
• Existing networks are ill-equipped to handle the extensive, uneven, and real-time traffic anticipated in the future.
• Cell site traffic exhibits complex temporal and spatial dependencies that are difficult to predict.

Purpose of the Study:

• To develop an adaptive graph convolutional network with gated recurrent unit (AGCN-GRU) for accurate cell site traffic prediction.
• To propose an AI-assisted intent-based traffic grooming scheme to manage unpredicted burst traffic.
• To optimize resource allocation, enhance utilization, and reduce delay and rejection ratios in 5G optical fronthaul networks.

Main Methods:

• Designed an AGCN-GRU network to learn temporal and spatial traffic dependencies, identifying spatial relations via traffic pattern similarity.
• Developed an AI-assisted intent-based traffic grooming scheme for automated cell site clustering and traffic grooming.
• Established a software-defined testbed for 5G optical fronthaul networks to deploy and evaluate the proposed schemes using real traffic datasets.

Main Results:

• The AGCN-GRU model accurately predicts traffic patterns by capturing temporal and spatial dependencies.
• The AI-assisted traffic grooming scheme effectively handles burst traffic and optimizes cell site clustering.
• Experimental results demonstrate significant improvements in network resource allocation and utilization.

Conclusions:

• The proposed AGCN-GRU and AI-assisted traffic grooming scheme effectively addresses the challenges in 5G optical fronthaul networks.
• The solution optimizes network resource allocation, leading to increased efficient resource utilization.
• The scheme successfully reduces average delay and rejection ratios, enhancing overall network performance.