UPF Reinforcement Learning Agent for Adaptive Bit Rate Video Shaping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Adaptive Bit Rate (ABR) shaping solutions in 5G telecommunications systems are inefficient due to reliance on static default shaping levels, which fail to adapt to changing network and video characteristics, leading to suboptimal video resolution selection and convergence issues, especially when different codecs are used.
Innovation Solution
Implementing a Reinforcement Learning Agent (RLA) within the User Plane Function (UPF) to dynamically determine and adjust video resolution shaping levels based on real-time network and video characteristics, eliminating the need for static configurations and enabling adaptation to individual user needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If static default shaping levels are used for video resolution shaping, then the system is simple to implement and operate, but it fails to adapt to changing network and video characteristics, leading to suboptimal video quality
Solution Approach 1:
The patent transforms the static default shaping level into a dynamic parameter that is continuously adjusted based on real-time network conditions and video characteristics. The shaping level is no longer fixed but adapts to changing bandwidth availability, network congestion, and video content requirements, resolving the contradiction between operational simplicity and adaptability.
Solution Approach 2:
The patent implements a feedback mechanism where the actual video transmission performance and network conditions are monitored and used to adjust the shaping level. This closed-loop control allows the system to automatically adapt to changing conditions while maintaining simple operation through automated decision-making based on observed performance metrics.
2Measurement precision
If resolution estimation requires multiple video chunks to be processed, then the estimation accuracy improves, but the time delay increases and default shaping levels become outdated
Solution Approach 1:
The patent applies preliminary shaping actions using available information before complete resolution estimation is available. Instead of waiting for multiple chunks to accumulate, the system applies initial shaping based on partial information and continuously refines it as more data becomes available, reducing the time delay while maintaining accuracy through iterative improvement.
Solution Approach 2:
The patent implements dynamic resolution estimation that updates continuously as video chunks are processed, rather than requiring a fixed number of chunks before estimation begins. This allows the shaping level to adapt progressively, reducing time delay while maintaining estimation accuracy through continuous refinement.
3Ease of manufacture
If heuristic algorithms are used for shaping level decisions, then the implementation is straightforward, but convergence issues occur and optimal resolutions are not achieved
Solution Approach 1:
The patent replaces heuristic algorithms with a more robust decision-making mechanism based on explicit performance metrics and optimization criteria. Instead of relying on rule-based heuristics that may fail to converge, the system uses measurable performance indicators and systematic optimization approaches that guarantee convergence to optimal or near-optimal solutions while maintaining implementation feasibility.
4Adaptability or versatility
If different video codecs are used, then video compatibility and versatility improve, but the relationship between bit rates and video resolutions becomes unpredictable, making static shaping ineffective
Solution Approach 1:
The patent changes the approach from relying on fixed bit rate to resolution mappings to dynamically adjusting shaping parameters based on actual video characteristics. The system monitors the relationship between bit rate and resolution for each specific video stream and codec combination, and adjusts the shaping level accordingly, making the system adaptable to different codecs while maintaining precise control over achieved resolution.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method of supporting Adaptive Bit Rate, ABR, video resolution shaping of a video data stream (32) of a video session transferred by a User Plane Function, UPF (10), in a Service Based Architecture, SBA,domain.The video resolution shaping is performed by the UPF (10) implementing a Reinforcement Learning Agent, RLA (25), operating with an observation space comprising a determined video resolution of a received video data stream (32), a reward space comprising a reward referring to a required video resolution, and an action space comprising video resolution shaping levels to be applied at the received video data stream (32). Complementary methods and devices for performing such a method in an SBA domain deployed in a telecommunications system are disclosed.