5G AI Model Transfer for AN Inference Format Compatibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in wireless communication systems is the efficient transfer and deployment of AI/ML models between network devices and user equipment, considering diverse network and UE deployments, individual AI capabilities, and varying levels of collaboration, which affects system performance and efficiency.
Innovation Solution
The implementation of AI model definition and transfer mechanisms that support one-sided and two-sided model transfers, utilizing the control and user planes of the 5G core network to facilitate model delivery to user equipment and base stations, ensuring compatibility and efficient inference operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI models are transferred between network devices and user equipment in wireless communication systems, then system performance and efficiency are improved, but device complexity and model format compatibility challenges increase
Solution Approach 1:
The patent transforms AI models from their original training formats into standardized communication formats suitable for wireless transmission. This involves changing parameters such as model representation (e.g., from raw neural network weights to compressed formats), data structure organization, and protocol encoding to ensure compatibility across different network devices and user equipment while maintaining model functionality and performance
Solution Approach 2:
The patent introduces standardized model transfer protocols and intermediate processing layers that act as mediators between AI model sources and target devices. These intermediaries handle format conversion, validation, and adaptation, allowing models to be transferred across diverse platforms without requiring each device to support every possible model format, thus reducing overall system complexity
2Adaptability or versatility
If AI model transfer mechanisms are implemented to support diverse network and UE deployments, then adaptability is improved, but overhead and resource consumption increase
Solution Approach 1:
The patent implements dynamic model transfer mechanisms that adapt to the specific capabilities, constraints, and requirements of different network devices and user equipment. The system can dynamically select appropriate transfer formats, compression levels, and protocol variations based on real-time conditions such as available bandwidth, device memory, processing power, and network state, thereby reducing unnecessary overhead while maintaining broad adaptability
Solution Approach 2:
The patent divides AI model transfer into segmented, modular components that can be selectively transferred based on device needs. Instead of transferring complete models regardless of requirement, the system segments models into essential and optional components, allowing receivers to obtain only what they need for their specific deployment scenario, thus reducing transfer overhead and resource consumption
Data Source
AI summary
An apparatus of a user equipment (UE), the apparatus comprising a processor, and a memory storing instructions that, when executed by the processor, configure the apparatus to receive, from a base station of an operator network, Protocol Data Units (PDUs) carrying Artificial Intelligence (AI) model data in either a control plane or a user plane, and decapsulate the PDUs to obtain and store the AI model data, wherein the AI model data is indicative of an AI model configured for inference in Access Network (AN) protocol layers at the UE.