Sparse ML Turbo Decoder for Low-Latency QoS-Adaptive Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional turbo decoders in base stations face high computational complexity, making them unsuitable for fast decoding in cloud-based Radio Access Networks (RANs) like VRAN, ORAN, and CRAN, which requires efficient data decoding within strict latency and delay constraints.
Innovation Solution
Implementing a Machine Learning (ML) based turbo decoder with sparse Deep Neural Network (DNN) or Convolutional Neural Network (CNN) decoders that determine optimal Modulation and Coding Scheme (MCS) and packet size, using a sparsity controller to adjust the neural network architecture based on Quality-of-Service (QoS) parameters, reducing computational complexity and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a conventional turbo decoder is used, then decoding accuracy is maintained, but computational complexity is high and latency is increased
Solution Approach 1:
The patent replaces the conventional mechanical/computational turbo decoder system with a Machine Learning-based decoder that uses trained neural networks. This substitution transforms the decoding process from iterative computational algorithms to direct ML model inference, significantly reducing computational complexity and latency while maintaining decoding accuracy.
Solution Approach 2:
The patent applies preliminary action by training the ML models offline beforehand. The decoding model is pre-trained on encoded data to learn optimal decoding patterns, so that during actual operation, the decoder can directly apply the learned patterns without performing complex real-time computations, thereby reducing operational complexity and latency.
2Device complexity
If sparsity is increased in ML models, then computational complexity is reduced, but decoding accuracy may deteriorate
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the sparsity parameter of the ML models based on QoS requirements. The sparsity controller modifies model parameters (such as pruning ratios or activation thresholds) to achieve the desired balance between computational complexity and decoding accuracy for different service scenarios.
Solution Approach 2:
The patent implements dynamics by making the ML model sparsity adjustable and adaptive rather than fixed. The sparsity controller dynamically modifies the model structure during operation based on real-time QoS parameters, allowing the system to optimize the trade-off between complexity and accuracy according to current network conditions and service requirements.
3Adaptability or versatility
If QoS parameters are considered for sparsity determination, then adaptability to different service requirements is improved, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary component called the sparsity controller that mediates between QoS parameters and ML model configuration. This intermediary translates high-level QoS requirements into specific model sparsity settings, simplifying the overall system architecture by centralizing the adaptation logic and avoiding direct complex interactions between multiple QoS parameters and model components.
Data Source
AI summary
A method for decoding data by an electronic device is provided. The method includes receiving, by the electronic device, encoded data, determining, by the electronic device, a sparsity of a plurality of Machine Learning (ML) models of a turbo decoder of the electronic device for decoding the encoded data based on Quality-of-Service (QoS) parameters, and decoding, by the electronic device, the encoded data using the turbo decoder based on the determined sparsity.