Cellular Network Quality Prediction Model Using Propagation Characteristics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting cellular network connectivity in road networks are imprecise, especially in areas without available measurements, and require expensive simulations or rely on few volunteer vehicle measurements.
Innovation Solution
A method that trains a prediction model using characteristics of radio signal propagation, including topographic elements, to predict cellular network quality at locations with no available measurements, using a multilayer perceptron neural network to correlate signal propagation characteristics with observed connection quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simulations are used to produce network coverage maps, then coverage areas can be identified, but the precision is too low for connected vehicles
Solution Approach 1:
The patent creates a predictive model that copies the relationship between propagation characteristics and connection quality from training data. Instead of running complex simulations, the system uses a trained model that replicates the behavior of the network, providing precise predictions without the computational burden of full simulations.
Solution Approach 2:
The patent transforms the problem from simulating entire network coverage to predicting connection quality based on specific propagation parameters. By focusing on key parameters (distance, obstacles, terrain) and using machine learning to process them, the system achieves high precision while reducing complexity.
2Loss of information
If volunteer vehicles are used to collect measurements, then real connection data can be obtained, but the coverage is insufficient for most road network
Solution Approach 1:
The system enables any vehicle to become a measurement point by collecting connection quality data and propagation characteristics. Each vehicle contributes to training the predictive model, transforming passive users into active participants that collectively build comprehensive road network coverage without requiring dedicated volunteer vehicles.
Solution Approach 2:
The patent collects and stores propagation characteristics and connection quality data in advance during normal vehicle operation. This preliminary data collection phase builds a training dataset that enables later predictions for any location, eliminating the need for extensive real-time measurement campaigns.
3Measurement precision
If traditional simulation methods are used, then network coverage can be estimated, but the cost and imprecision are high
Solution Approach 1:
The patent creates a lightweight predictive model that copies the essential relationships from training data. This model provides accurate connection quality predictions without requiring expensive and resource-intensive simulations, achieving high precision with minimal computational resources.
Data Source
AI summary
A method for predicting a value representative of a quality of connection of a vehicle to a cellular network from a target location on a road network including: training a prediction model on the basis of a determined characteristic of propagation between an access point and the position of a training vehicle and on the basis of a value representative of a quality of the connection of the vehicle to said access point; and predicting a second value representative of a quality of connection to a cellular network at the target location on a road network on the basis of a second characteristic of propagation of a radio signal between the target location and an identified target access point by applying the trained model.