Machine Learning Geolocation for Cellular Coverage Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network monitoring methods for cellular communication systems are time-consuming, expensive, and do not accurately represent the experience of real users, particularly in determining coverage problems such as coverage holes and weak coverage, which are often caused by physical obstructions or inadequate RF planning.
Innovation Solution
A machine learning method is employed by network monitoring devices to determine the geolocation of user equipment (UE) in cellular communication systems by analyzing radio signal attributes, including Timing Advance (TA) values and GPS data, to accurately identify coverage issues and optimize network coverage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional drive tests are used to measure network coverage, then geographic positional information can be obtained, but the process is time-consuming and expensive
Solution Approach 1:
The patent creates virtual copies of drive test functionality by using UE-based measurements and machine learning models to simulate traditional drive test data collection. The system processes signaling records from regular UEs to generate coverage maps, replacing the need for physical drive test vehicles while maintaining measurement accuracy.
Solution Approach 2:
The patent replaces the mechanical drive test system (physical vehicles, scanners, and technicians) with an automated electronic system that uses machine learning algorithms to process signaling records from user devices. This substitution eliminates the need for physical movement and manual data collection while achieving the same coverage measurement objectives.
2Measurement precision
If traditional drive tests are used to measure network coverage, then coverage data can be collected, but operator costs increase
Solution Approach 1:
The patent implements a self-service mechanism where regular user equipment devices automatically provide measurement data for network coverage optimization. UEs contribute their own signaling records and location information to the system, eliminating the need for dedicated drive test resources and reducing operator expenditures on manual coverage surveys.
Solution Approach 2:
The patent makes the signaling record processing system multi-functional by using the same infrastructure to serve both regular communication purposes and network optimization purposes. The machine learning model processes various types of signaling records for multiple objectives including coverage mapping, interference detection, and capacity planning, maximizing resource utilization.
3Measurement precision
If drive tests are conducted from vehicles, then road-based coverage data is obtained, but pedestrian user experience is not accurately represented
Solution Approach 1:
The patent changes the measurement parameters by collecting data from stationary or slowly moving UEs in various locations including buildings and pedestrian areas, rather than from moving vehicles on roads. This parameter change in measurement position and movement state enables accurate representation of pedestrian user experience and indoor coverage conditions.
Solution Approach 2:
The patent collects excessive measurement data from multiple UEs in various locations and uses machine learning to extract the relevant patterns. By gathering more data points from diverse positions and conditions than traditional drive tests, the system ensures comprehensive coverage of pedestrian experiences and identifies coverage issues that would be missed by road-based measurements.
Data Source
AI summary
A machine learning method performed by a communication network monitoring device in which an incoming signaling record is received that includes radio signal attributes from a UE in the cellular communication network. A determination is made as to whether the UE incoming signaling record contains location (GPS) data. If the UE incoming signaling record contains GPS data, a machine learning model is generated for determining a location of future UEs in the communication network utilizing the GPS data and the radio signal attributes from the incoming UE signaling record. And if GPS data is not included in the UE incoming signaling record, then first a corrected TA value is determined which is then used, along with other radio signal attributes of the UE, to determine/predict a geolocation for the UE using machine learning techniques.