DCI Trigger for Combined ML Model Configuration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communication systems, particularly in 5G NR, face challenges in efficiently configuring machine learning (ML) models at user equipment (UE) due to limited physical downlink control channel (PDCCH) resources, necessitating a method to balance ML model configuration time with resource costs.
Innovation Solution
The proposed solution involves using downlink control information (DCI) to trigger or determine the configuration of ML models by associating backbone/general blocks with specific/dedicated blocks, allowing for the generation of combined ML models that reduce signaling costs and enhance flexibility for different tasks/conditions, with the DCI indicating the association between these blocks to configure the ML model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ML model configuration is performed using traditional signaling methods, then the ML model can be configured at the UE, but the PDCCH resource usage increases and configuration time is extended
Solution Approach 1:
The ML model is segmented into a backbone block and multiple task-specific blocks. The backbone block is configured once and reused across multiple tasks, while only the necessary task-specific blocks are configured via DCI. This segmentation reduces the overall configuration time and PDCCH resource usage while maintaining complete ML model functionality when needed.
Solution Approach 2:
The system dynamically switches between using only the backbone block for low-latency tasks and combining the backbone block with task-specific blocks for enhanced performance tasks. This dynamic adaptation allows the system to optimize configuration time based on task requirements, reducing average configuration time while maintaining reliability.
2Reliability
If traditional signaling methods are used for ML model configuration, then complete ML models can be configured, but PDCCH resource consumption increases
Solution Approach 1:
The ML model configuration is segmented into a backbone block configured once and multiple task-specific blocks configured selectively. This reduces the total amount of signaling required while ensuring complete model configuration is achieved when task-specific blocks are combined with the backbone block.
Solution Approach 2:
The backbone block serves as a universal foundation that can be combined with different task-specific blocks for multiple tasks. This multi-functionality reduces PDCCH resource usage by avoiding redundant transmission of common model parameters across different tasks.
3Adaptability or versatility
If full ML models are configured for all tasks, then task performance is optimized, but configuration complexity and time increase
Solution Approach 1:
The ML model is divided into a shared backbone block and task-specific blocks. This segmentation allows the system to manage complexity by configuring only the necessary task-specific blocks for each task while reusing the backbone block, thereby reducing overall configuration complexity while maintaining task-specific adaptability.
Solution Approach 2:
The system dynamically selects which task-specific blocks to combine with the backbone block based on the current task requirements. This dynamic approach simplifies the configuration process by avoiding the need to pre-configure all possible task combinations, reducing device complexity while maintaining versatility.
Data Source
AI summary
A base station may set one or more bits of DCI that at least one of indicate or trigger a configuration of an ML model at a UE. The configuration may be based on an association between at least one first ML block for a first procedure and at least one second ML block for a second procedure. The at least one second ML block may be dedicated to a task included in a plurality of tasks associated with the at least one first ML block. The base station may transmit the DCI including the one or more bits to the UE, which may cause the UE to configure the ML model including the association between the at least one first ML block for the first procedure and the at least one second ML block for the second procedure.