PHY Assistance Signaling for Adaptive AI Inference Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The integration of AI/ML models in wireless communication networks faces challenges due to varying processing times across different devices, with current standards defining worst-case performance, which prevents faster devices from benefiting from reduced latency.
Innovation Solution
A method to determine and signal the inference time for AI/ML models in wireless communication networks, allowing devices to indicate their processing capabilities and requirements, enabling efficient use of AI/ML models based on their specific hardware and capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If worst-case processing time is used for AI/ML model execution, then all devices can operate with a unified time parameter, but faster devices cannot benefit from reduced latency
Solution Approach 1:
The patent introduces dynamic processing time parameters that adapt to individual device capabilities. Instead of using a fixed worst-case time parameter for all devices, the system dynamically determines and signals appropriate processing times based on each device's actual AI/ML performance characteristics, allowing faster devices to utilize shorter processing times for reduced latency.
Solution Approach 2:
The patent changes the parameter of processing time from a static worst-case value to a dynamic value that varies based on device capabilities. The network determines and signals different processing time parameters to different devices based on their inferred AI/ML performance, enabling parameter optimization for each device's specific hardware capabilities.
2Productivity
If AI/ML models are integrated into wireless communication networks, then enhanced functionality and performance are achieved, but varying processing times across different devices create complexity
Solution Approach 1:
The patent implements a feedback mechanism where devices signal their AI/ML processing capabilities and performance characteristics to the network. The network uses this feedback information to determine appropriate processing time parameters for each device, creating a closed-loop system that manages device diversity without increasing overall system complexity.
Solution Approach 2:
The patent enables devices to self-report their AI/ML processing capabilities and performance characteristics. Each device autonomously determines and signals its own processing time requirements based on its hardware capabilities, eliminating the need for network-side complexity in characterizing each device's performance.
3Speed
If processing time is optimized for faster devices, then latency is reduced, but slower devices may not meet minimum performance requirements
Solution Approach 1:
The patent performs preliminary characterization of device AI/ML processing capabilities before actual model execution. Devices signal their processing capabilities in advance, allowing the network to pre-determine appropriate processing time parameters that guarantee minimum performance requirements while optimizing for faster devices.
Data Source
AI summary
Embodiments provide an apparatus of a wireless communication network, the wireless communication network using one or more Artificial Intelligence/Machine Learning, AI/ML, models for one or more use cases, wherein the apparatus is to determine an inference time for one or more of the AI/ML models to be used in one or more network entities of the wireless communication network.