Communication Network Node VAE for Proactive Radio Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication network techniques face challenges in efficiently managing radio resource parameters for load balancing, carrier aggregation, and dual connectivity due to inefficient UE selection and resource wastage, particularly in 5G networks with diverse and UE-specific traffic patterns, leading to suboptimal throughput and resource drain.
Innovation Solution
Utilizing a Variational Autoencoder (VAE) trained on historical data flows to predict traffic distributions for nodes in the network, enabling early configuration of radio resource parameters such as carrier aggregation, dual connectivity, and load balancing without the need for observation periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional load balancing techniques are used to balance traffic between cells, then network capacity is increased, but many UEs perform unnecessary inter-frequency measurements wasting signalling resources, processing power, and battery energy
Solution Approach 1:
The patent applies preliminary action by training machine learning models (neural networks) on historical traffic data beforehand to predict future traffic patterns. This allows the system to proactively identify UEs likely to benefit from load balancing before initiating inter-frequency measurements, rather than reacting after measurements are already performed. The prediction model is trained in advance using past traffic flow data, enabling early identification of suitable candidates for cell transfer.
Solution Approach 2:
The system implements self-service by using automatically generated traffic predictions from machine learning models to drive load balancing decisions. Instead of relying on manual configuration or simple threshold-based rules, the network autonomously predicts which UEs will benefit from load balancing based on their historical traffic patterns, and automatically initiates appropriate actions. The system serves itself by generating its own intelligence from operational data.
2Productivity
If carrier aggregation is configured for all UEs to increase throughput, then network performance is improved, but signalling overhead and device resources are unnecessarily drained for UEs with small traffic packets
Solution Approach 1:
The patent applies local quality by tailoring carrier aggregation configuration to individual UE traffic characteristics rather than applying a uniform configuration to all UEs. The machine learning model predicts traffic patterns for each UE specifically, and based on these localized predictions, the system selectively enables or disables carrier aggregation for each device. This ensures that UEs with small traffic packets do not receive unnecessary carrier aggregation configurations, while UEs with large traffic demands do receive the benefit.
3Manufacturing precision
If observation periods are used to determine traffic patterns before configuring radio resource parameters, then accurate configuration is achieved, but time is lost and throughput is reduced during the observation period
Solution Approach 1:
The patent resolves this contradiction by performing the analytical work in advance through offline training of machine learning models on historical traffic data. Instead of observing traffic patterns in real-time before configuration (which causes delay), the system has already learned traffic patterns from past data during model training. When a UE connects, the pre-trained model can immediately generate accurate traffic predictions without requiring any observation period, thus eliminating the time loss while maintaining configuration accuracy.
4Reliability
If dual connectivity is established to enable user plane connections via multiple radio nodes, then network capacity and reliability are improved, but setup complexity and resource consumption increase
Solution Approach 1:
The system applies preliminary action by using pre-trained machine learning models to predict which UEs are suitable candidates for dual connectivity before attempting to establish it. The model analyzes historical traffic patterns and predicts future traffic demands, identifying UEs that will benefit from dual connectivity in advance. This allows the network to proactively prepare and configure dual connectivity for selected UEs, reducing setup complexity by avoiding attempts to establish dual connectivity with UEs that would not benefit from it.
Data Source
AI summary
A method for managing a first node in a communication network is disclosed, wherein the first node is operable to exchange traffic flows with other nodes in the communication network. The method includes using a Variational Autoencoder (VAE) to generate a predicted traffic distribution for the first node, wherein the VAE has been trained using information about historical data flows exchanged by the first node with at least one other node in the communication network, and configuring at least one radio resource parameter of the first node based on the obtained predicted traffic distribution for the first node. Also disclosed are a method including training a VAE and nodes and a computer program product suitable for carrying out such methods.