Cellular Network Performance Estimator Using Graph Autoencoders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network simulators for cellular networks are difficult to calibrate and require extensive real-world data sets for training, while AI models need costly data collection to predict performance changes, making network optimization challenging.
Innovation Solution
A performance estimator model using a combination of autoencoders and graph neural networks (GNN) that estimates network performance without requiring data sets of pre-and post-change measurements, leveraging network snapshots with varying configuration parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional network simulators are used to model network elements and estimate performance, then performance prediction capability is provided, but the system becomes difficult to calibrate and requires extensive configuration parameters
Solution Approach 1:
The patent replaces traditional mechanical simulation approaches (network simulators that model physical network elements like antennas, links, and propagation) with an AI-based neural network model. This substitution eliminates the need for complex calibration of physical parameters while maintaining performance prediction capability, directly resolving the contradiction between measurement precision and device complexity
Solution Approach 2:
Instead of creating a detailed digital twin that copies all network elements and their interactions, the patent uses a simplified neural network model that learns performance patterns from training data. This copying approach captures essential performance characteristics without replicating the full complexity of the network, reducing calibration difficulty while preserving prediction accuracy
2Measurement precision
If AI models are trained using real data from configuration parameter modifications, then performance estimation accuracy is improved, but data collection cost and time increase significantly
Solution Approach 1:
The patent performs preliminary action by training the AI model on historical network data that already exists, rather than requiring new data collection for each optimization scenario. The model is pre-trained on diverse configuration scenarios, enabling it to generalize to new situations without requiring additional real-world experiments, thus reducing time loss while maintaining estimation accuracy
Solution Approach 2:
The trained AI model serves multiple functions: it can estimate performance for any configuration parameter change, work across different network scenarios, and provide predictions without requiring additional data collection. This universality allows a single training phase to support numerous optimization tasks, eliminating the need for repeated data collection cycles
3Manufacturing precision
If comprehensive network modeling is performed to account for all configuration parameters, then optimization accuracy is improved, but computational complexity and resource requirements increase
Solution Approach 1:
The patent extracts only the essential performance-related features from the comprehensive set of configuration parameters, rather than modeling all parameters in detail. The neural network focuses on learning the relationship between key configuration changes and performance outcomes, filtering out unnecessary complexity while maintaining optimization accuracy
Solution Approach 2:
The patent transforms the complex multi-parameter optimization problem into a simpler parameter change detection problem. The AI model learns to predict performance changes based on configuration parameter modifications, working with parameter deltas rather than absolute values, which reduces model complexity while preserving optimization precision
Data Source
AI summary
According to an aspect, there is provided a comput-er-implemented method for training a performance estimator model (22) for estimating the performance of a cellular network. The performance estimator model (22) comprises an encoder stage (24) comprisesing a plurality of encoders and a decoder stage (26) comprising a plurality of decoders. The network comprises a plurality of cells, and the cellular network has a plurality of configuration parameters and a configuration parameter of interest that are each configurable per cell. The method comprises: (I) obtaining (901) a training data set, the training data set comprising measurements of a plurality of performance parameters for the plurality of cells, wherein different values of the configuration parameters are being used among the plurality of cells, wherein the training data set further comprises respective values of the plurality of configuration parameters for the plurality of cells; (II) training (903) the encoders to encode the training data set into a respective representation for each cell, wherein each encoder receives, for a respective cell, the measurements of the plurality of performance parameters and corresponding values of the plurality of configuration parameters for that cell, and wherein a layer of each of the plurality of encoders are interconnected as a neural network representing the cellular network such that information on relationships between different pairs of cells in the cellular network is taken into account in the encoding; (ill) training (905) the decoders to decode a respective representation to determine a subset of the plurality of performance parameters for the respective cell associated with the configuration parameter of interest, wherein each decoder receives the respective representation and a current value of the configuration parameter of interest, wherein a layer of each of the plurality of decoders are interconnected as a neural network representing the cellular network such that information on relationships between different pairs of cells in the cellular network is taken into account in the decoding; (iv) determining (907) a value of a loss metric that is based on a difference between the input training data set and the output of the decoders; and (v) repeating (909) steps (II), (ill) and (iv) to retrain the encoders and decoders to obtain an improved value of the loss metric.