PHY Processing Time Prediction for MAC Scheduler Utilization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication networks face challenges in accurately predicting physical layer processing times due to the complexity of the PHY layer, leading to inefficiencies in resource allocation and potential disruptions from unpredictable processing spikes.
Innovation Solution
Implementing an AI/ML model trained to predict PHY layer processing times based on configuration data, accounting for previously scheduled tasks and system constraints, allowing for optimized task execution sequences and efficient resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a safety margin is maintained to keep processing load below the hard limit, then system reliability is improved, but resource utilization deteriorates
Solution Approach 1:
The system performs preliminary actions by training an AI/ML model offline to predict PHY layer processing times. This pre-computed knowledge enables the MAC scheduler to make informed decisions without needing to maintain large safety margins, thus improving resource utilization while maintaining reliability.
Solution Approach 2:
The system implements feedback by using the AI/ML model to continuously predict processing times based on configuration data and previously scheduled tasks. This feedback loop allows dynamic adjustment of scheduling decisions, enabling tighter load management without compromising system reliability.
2Productivity
If the physical layer complexity is increased to handle more tasks, then productivity is improved, but the difficulty of detecting and measuring processing time deteriorates
Solution Approach 1:
The AI/ML model serves as an intermediary between the complex PHY layer and the MAC scheduler. It translates complex processing configurations into accurate processing time predictions, making the complex system measurable and manageable without reducing PHY layer capability.
Solution Approach 2:
The system creates a virtual model (AI/ML prediction model) that replicates the behavior of the physical layer processing. This copy allows accurate measurement and prediction of processing times without directly measuring the complex physical layer operations, thus maintaining productivity while improving measurability.
3Productivity
If resource utilization is increased to reduce idle time, then productivity is improved, but system stability deteriorates due to unpredictable processing spikes
Solution Approach 1:
The system performs preliminary analysis by training the AI/ML model on historical processing data to learn patterns and predict future processing times. This pre-computed knowledge allows the system to operate closer to capacity limits without sacrificing stability, as the scheduler can anticipate and prepare for processing variations.
Solution Approach 2:
The system implements dynamic scheduling by using AI/ML predictions to adaptively adjust task allocation based on current and predicted processing loads. This dynamic approach allows the system to maintain high utilization while responding flexibly to processing spikes, thus improving productivity without compromising stability.
Data Source
AI summary
The technology described herein is directed towards having a trained artificial intelligence/machine learning (AI/ML) model provide an accurate physical layer (PHY/L1) processing time prediction that a MAC (medium access control) scheduler can use to schedule resources more efficiently. The AI/ML model is trained with large, easily obtainable datasets of configurations of tasks, to return an accurate processing time estimate based on input configuration data representing task data to be scheduled. The model can map a sequence of configurations for consecutive time slots to a sequence of expected processing times for each slot, while capturing temporal information for previously scheduled but not yet processed tasks, making it possible to optimize task execution sequences by choosing an optimal permutation for configurations/tasks. A real time scheduler can test multiple task hypotheses to select the most optimal one for a time slot.