Twin Agent Neural Network Routing for Dynamic Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data traffic routing systems in communication networks face challenges in efficiently routing data packets due to variations in latency and bandwidth, particularly in multi-access architectures, as current methods like MPTCP are not sophisticated enough to handle these changes effectively, leading to suboptimal performance.
Innovation Solution
A method involving twin agents, where a first agent generates routing instructions using a static current routing model and sends experience information to a second agent to train a neural network when a threshold is exceeded, updating the model to provide accurate and efficient routing instructions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If packet scheduling algorithms like MPTCP are used for path selection, then routing decisions can be made based on available parameters, but the algorithms are not sophisticated enough to accurately model network variations in latency and bandwidth
Solution Approach 1:
The patent replaces traditional mechanical packet scheduling algorithms with a neural network-based intelligent system. The neural network learns complex network performance patterns from historical data and provides accurate predictions of latency and bandwidth, overcoming the limitations of rule-based algorithms while adapting to dynamic network conditions.
Solution Approach 2:
The patent introduces a neural network as an intermediary between network status monitoring and routing decision-making. This intermediary processes raw network parameters, learns complex relationships, and outputs optimized routing decisions, bridging the gap between simple parameter collection and sophisticated routing control.
2Reliability
If neural network training is performed continuously to maintain model accuracy, then routing performance improves, but computational resources and time are consumed
Solution Approach 1:
The patent implements periodic training of the neural network at predetermined intervals rather than continuous training. This approach maintains model accuracy by periodically updating the neural network with new network data, while avoiding the excessive computational overhead and time consumption of continuous training, thus balancing reliability with time efficiency.
3Productivity
If multiple access networks are aggregated to increase bandwidth and reliability, then network performance improves, but determining the best path for each data packet becomes more complex
Solution Approach 1:
The patent uses the neural network to create optimized routing decisions based on learned patterns from historical network data. Instead of complex real-time analysis of all possible paths, the neural network provides copied successful routing strategies that have proven effective in similar network conditions, simplifying the path selection process while maintaining high throughput.
Data Source
AI summary
A data traffic routing method and apparatus for controlling data traffic in a communication network, the method comprising: receiving, at a first agent from a User Plane Function, communication network status information; calculating, by the first agent, data traffic routing instructions using a current routing model; sending by the first agent: the data traffic routing instructions to the User Plane Function; and experience information to a second agent; storing, at the second agent, the experience information; determining, at the second agent, if the number of instances of stored experience information exceeds a predetermined threshold; and if it is determined that the number of instances of stored experience information exceeds a predetermined threshold: training a neural network using the instances of stored experience information; and updating the current routing model using results of the neural network training.