AI Wireless Network Management Through Cross-Band Signal Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting signal propagation in new frequency bands, such as UHF, SHF, and EHF, are cumbersome, costly, and lack accuracy due to the need for extensive measurement campaigns and outdated databases, failing to adapt to environmental changes.
Innovation Solution
A machine-learning model, such as a Deep Neural Network (DNN), is trained on reference signal strength maps from multiple frequency bands to predict deviations in signal strengths, allowing for the estimation of target signal strength maps in new frequency bands using source signal strength maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If propagation model optimization techniques are used to characterize wireless network propagation environments, then signal propagation accuracy can be improved, but extensive measurement campaigns are required which increases cost and time
Solution Approach 1:
The patent creates a virtual copy of the physical propagation environment by training a machine learning model on measured data from existing frequency bands. This virtual model replicates propagation characteristics without requiring physical measurement campaigns in new frequency bands, thereby achieving accuracy while reducing time and cost.
Solution Approach 2:
The patent performs preliminary training of the machine learning model using measurement data collected in advance from existing frequency bands. This preliminary action creates a ready-to-use propagation model that can predict characteristics in new frequency bands without requiring new measurement campaigns, thus resolving the time-loss contradiction.
2Measurement precision
If ray tracing techniques are used to simulate radio wave propagation, then propagation characteristics can be predicted, but detailed 3D databases and material descriptions are required which increase complexity and cost
Solution Approach 1:
Instead of using complex ray tracing simulations that require detailed 3D databases, the patent creates a simplified virtual model by copying propagation patterns from existing frequency bands. This machine learning-based approach replicates propagation behavior without requiring complex geometric databases or material property descriptions, thereby reducing system complexity while maintaining prediction capability.
3Measurement precision
If independent propagation models are adjusted per cell to improve local adaptation, then local accuracy can be improved, but the cost and complexity increase significantly
Solution Approach 1:
The patent develops a universal machine learning model that can be applied across multiple cells and frequency bands. This single model performs the function of multiple cell-specific models by learning general propagation patterns that can be adapted to different locations through the frequency band mapping approach, thereby reducing model management complexity while maintaining local adaptation capability.
4Measurement precision
If propagation models are tuned using massive measurement campaigns to improve accuracy, then model precision can be improved, but the cost and resource requirements increase
Solution Approach 1:
The patent copies propagation characteristics from existing frequency bands to new frequency bands using machine learning. Instead of collecting massive measurement data for each new frequency band, the system replicates propagation patterns from already-measured bands, thereby achieving model accuracy without requiring proportional increases in measurement data volume.
Solution Approach 2:
The patent changes the frequency band parameter as the key variable for model adaptation. Rather than collecting new measurements for each scenario, the system adjusts predictions by learning frequency band-specific deviations from a base model trained on existing data, thereby achieving accuracy with reduced measurement requirements.
5Loss of information
If periodical measurement reports are activated to obtain training data, then model training data availability is improved, but network load and signaling overhead increase
Solution Approach 1:
The patent merges the training data collection process with existing network operations by utilizing measurement data from existing frequency bands that are already being monitored for network optimization. This combining approach obtains necessary training data without activating separate periodical measurement reporting mechanisms, thereby avoiding additional signaling overhead and energy consumption.
Data Source
AI summary
Deviations of signal strengths in a first frequency band from signal strengths in at least one second frequency band are predicted based on a trained machine-learning model (350′). At least one source signal strength map is obtained. The at least one source signal strength map describes signal strengths in the at least one second frequency band for a coverage area of the wireless communication network. Based on the at least one source signal strength map and the predicted deviations of signal strengths, at least one target signal strength map describing signal propagation in the first frequency band for the coverage area is determined.