POC Spectrum Sharing Using CNN Prediction and RL Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dynamic spectrum management (DSM) systems face challenges in efficiently utilizing partially overlapping channels (POCs) for concurrent spectrum access, particularly for hardware-constrained Internet-of-Things devices, due to complex power control and interference cancellation algorithms, and the dynamic nature of IoT device environments.
Innovation Solution
A deep learning method and system that utilizes a channel interleaving technology, combining a convolutional neural network for channel state prediction and reinforcement learning to optimize channel allocation strategies, maximizing throughput by learning from historical channel state information and using coding redundancy for data recovery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional power control and interference cancellation algorithms are used for POC allocation, then spectral efficiency can be improved, but device complexity and computational overhead increase significantly
Solution Approach 1:
The patent replaces conventional mechanical/mathematical algorithms (power control and interference cancellation) with an artificial intelligence-based system. The AI model learns optimal POC allocation strategies through training on historical channel state information, substituting complex real-time calculations with pre-learned decision patterns that can be executed with lower computational overhead on IoT devices.
2Measurement precision
If complete and accurate channel information is measured for POC allocation, then allocation accuracy is improved, but measurement difficulty and time consumption increase
Solution Approach 1:
The patent collects and stores historical channel state information in advance before actual POC allocation is needed. This historical data serves as pre-prepared training material for the AI model, eliminating the need to measure complete channel information in real-time during allocation decisions. The AI learns from this pre-collected data to make accurate allocation predictions.
Solution Approach 2:
Instead of measuring complete and accurate channel information directly during operation, the system uses historical channel state information as a copy or representation of the actual channel conditions. The AI model learns to map these historical patterns to optimal allocation strategies, avoiding the need for precise real-time measurement while maintaining effective allocation performance.
3Device complexity
If static spectrum allocation strategy is used, then system simplicity is maintained, but spectrum utilization efficiency deteriorates
Solution Approach 1:
The patent transitions from static spectrum allocation to dynamic spectrum allocation by implementing an AI-based system that adapts to changing channel conditions. The AI model continuously learns from historical channel state information and adjusts POC allocation strategies dynamically, allowing the system to respond to environmental changes while maintaining relative simplicity through automated decision-making.
Solution Approach 2:
The AI-based POC allocation system is self-learning and self-adjusting, automatically improving spectrum utilization without requiring complex external control mechanisms. The system serves itself by using its own historical performance data to train and refine its allocation strategies, eliminating the need for manual configuration or complex centralized control while achieving high spectral efficiency.
Data Source
AI summary
The present application discloses a deep learning method and system for spectrum sharing of partially overlapping channels. The method comprises the following steps: in response to a received user transmission request, inputting, by a base station, channel state information CSI of a plurality of historical time slices into a trained channel prediction convolutional neural network model and outputting predicted channel state information CSI of a next time slice; and inputting the channel state information CSI of the next time slice into a reinforcement learning model and obtaining a channel allocation strategy of each user equipment in a collision domain of the base station so as to realize a maximum throughput of simultaneous communication of each user equipment, wherein the reinforcement learning model is obtained by training by taking bandwidth efficiency performance as a reward. The present application is high in universality, bandwidth utilization rate and throughput for communication networks.