PDCP PDU Size Determination Using ML-Based SDU Concatenation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face inefficiencies in processing and transmitting service data units due to overheads such as processing time, security key setup, and heterogeneous payloads, which can be addressed by concatenating multiple service data units into a single concatenated service data unit using a machine learning model.
Innovation Solution
A UE implements a machine learning model to determine a concatenation policy for service data units at the PDCP layer, generating a concatenated service data unit based on radio conditions, traffic type, and UE internal conditions, thereby optimizing the concatenation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple service data units are transmitted individually, then processing and security key setup overhead increases, but transmission simplicity is maintained
Solution Approach 1:
The patent merges multiple service data units into a single concatenated service data unit at the PDCP layer. This combining approach reduces the number of individual processing operations and security key setups required, directly improving processing efficiency while maintaining transmission functionality
Solution Approach 2:
The patent implements dynamic concatenation policies that adapt to varying radio conditions, traffic types, and UE internal states. The machine learning model continuously adjusts concatenation parameters based on current system conditions, optimizing the balance between processing efficiency gains and policy complexity management
2Loss of time
If service data units are concatenated into a single unit, then processing time and overhead are reduced, but determination of optimal concatenation size becomes more complex
Solution Approach 1:
The patent employs a machine learning model deployed at the UE that autonomously determines optimal concatenation sizes based on observed radio conditions, traffic patterns, and performance feedback. The system self-optimizes without requiring complex external control mechanisms, reducing processing time while managing the complexity of size determination through intelligent automation
Solution Approach 2:
The patent implements feedback mechanisms where the UE monitors the effects of concatenation decisions on transmission performance and uses this information to refine future concatenation size choices. The machine learning model learns from past performance data to progressively improve its ability to determine optimal concatenation sizes, reducing processing time over the system's operational life
3Use of energy by moving object
If concatenation policy is determined by machine learning model, then processing efficiency and power consumption are improved, but model training and implementation complexity increases
Solution Approach 1:
The patent extracts the complex machine learning model operations from the main transmission path, implementing them as separate training and inference phases. The model is trained offline or during low-activity periods and then deployed for rapid inference during active transmission, reducing real-time processing complexity while maintaining the energy efficiency benefits of intelligent concatenation decisions
Solution Approach 2:
The patent performs machine learning model training and optimization in advance, before actual data transmission begins. The model is pre-trained on historical data and fine-tuned based on initial system observations, so that during normal operation, only lightweight inference is required. This preliminary action reduces the computational burden during power-intensive transmission periods while achieving energy efficiency through optimized concatenation policies
Data Source
AI summary
Methods, systems, and devices for method for wireless communication are described. A user equipment (UE) may receive, at a packet data convergence protocol layer of the UE, a set of service data units for an uplink transmission. Th UE may then concatenate the set of service data units to generate a concatenated service data unit in accordance with a concatenation policy generated by a machine learning model. The machine learning model may be based on a radio condition, a traffic type associated with the set of service data units, an internal condition associated with the UE, a network configuration including a range threshold or a performance threshold or both, or any combination thereof. The UE may then transmit the uplink transmission including the concatenated service data unit.