Wireless Network Signaling Framework for Adaptive ML Data Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing air interface measurement collection frameworks, such as minimization of drive tests (MDT), are not optimized for artificial intelligence or machine learning applications, lacking flexibility and requiring significant over-collection and processing, and are not adaptable to different scenarios and use cases.
Innovation Solution
A protocol and signaling framework is developed to enable machine learning models in wireless communication networks, allowing for configurable data collection parameters, real-time or near-real-time data gathering, and optimized data collection for machine learning model training, with support for lifecycle management of ML models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional MDT frameworks are used for data collection, then network measurement capability is provided, but flexibility and adaptability to different scenarios are reduced
Solution Approach 1:
The patent implements dynamic configurability of data collection parameters through signaling frameworks that allow network entities to adjust measurement parameters in real-time based on different scenarios and use cases, transforming the static MDT framework into a dynamic adaptive system
Solution Approach 2:
The patent enables parameter reconfigurations by introducing signaling mechanisms that dynamically modify data collection parameters such as measurement types, reporting intervals, and target entities, allowing the system to adapt to varying network conditions and ML training requirements without redesigning the entire framework
2Productivity
If conventional MDT frameworks are used for data collection, then network measurement capability is provided, but significant over-collection and subsequent filtering and processing are required
Solution Approach 1:
The patent applies preliminary action by configuring data collection parameters in advance based on specific ML training requirements and scenario needs, so that only relevant data is collected from the outset, eliminating the need for subsequent filtering and processing of unnecessary data
Solution Approach 2:
The patent extracts only the necessary data elements required for specific ML applications by implementing targeted data collection configurations, removing unnecessary data gathering steps and reducing the burden of post-collection filtering and processing
3Reliability
If frequent and dynamic parameter reconfigurations are implemented, then relevant information is obtained for model retraining, but system complexity increases
Solution Approach 1:
The patent implements feedback mechanisms where ML model performance and network conditions are continuously monitored, and this feedback drives automated parameter reconfigurations, reducing the complexity of manual reconfiguration while ensuring model training accuracy through data-driven adjustments
Data Source
AI summary
Aspects of the subject disclosure may be directed to, for example, a method including determining a set of data collection parameters that are configurable to collect data indicative of network events occurring in real time or near real time in the wireless communication networks, receiving the collected data based on the set of data collection parameters from a group of network entities operating in the wireless communication networks, based on the received collected data, and generating training data for a machine learning model deployed in the wireless communication networks. Other embodiments are disclosed.