AI/ML Positioning Dataset Indexing for Consistent Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a need for improved AI/ML positioning protocols in wireless communication systems, particularly in 5G NR, to enhance the accuracy and efficiency of location services by enabling network entities to select suitable AI/ML models based on current positioning configurations and radio characteristics.
Innovation Solution
A method for transmitting and receiving dataset identifiers to index AI/ML models related to positioning, allowing network entities to store and log corresponding positioning configurations and radio statistics, thereby training models that match the current network conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If network entities share proprietary configuration changes to improve AI/ML model training accuracy, then positioning accuracy is improved, but information security and proprietary protection deteriorate
Solution Approach 1:
The patent introduces a location server as an intermediary that collects positioning configurations from multiple network entities (gNodeBs, eNodeBs) and provides indexed datasets to UEs without exposing the proprietary configuration details. The location server acts as a mediator that enables AI/ML model training while protecting proprietary information through data indexing and selective information sharing.
Solution Approach 2:
The patent creates indexed copies of positioning configuration data that can be referenced without exposing the original proprietary configurations. By indexing datasets based on configuration parameters, the system allows network entities to share training data through references rather than direct data exchange, maintaining information security while enabling accurate model training.
2Productivity
If network entities collect and store positioning configurations and radio statistics for AI/ML model training, then model training performance is improved, but system complexity and data management burden increase
Solution Approach 1:
The patent segments the data collection and management function by introducing a location server that centralizes the indexing of positioning configurations and radio statistics. Each network entity collects only local data and stores it with a unique index, while the location server manages the global indexing structure. This segmentation reduces the data management burden on individual network entities while maintaining comprehensive training data availability.
Solution Approach 2:
The patent implements preliminary indexing of positioning configurations and radio statistics before AI/ML model training. By pre-organizing data into indexed datasets at the location server, the system prepares training data in advance, reducing the complexity of data retrieval and management during the training process and improving overall training efficiency.
Data Source
AI summary
Aspects presented herein may enable a consistency between multiple network entities in artificial intelligence (AI) or machine learning (ML) (AI/ML) related positioning training and inference. In one aspect, a first network entity transmits, to a second network entity, a request for an identifier (ID) to be used for indexing a set of datasets associated with at least one AI/ML model related to positioning. The first network entity receives, from the second network entity based on the request, the ID to be used for indexing the set of datasets associated with the at least one AI/ML model related to positioning. The first network entity stores, based on the ID, at least one of a set of positioning configurations or a set of radio statistics associated with the first network entity. The first network entity indexes the set of datasets with the ID.