User Equipment Reporting Computation Capability for AI/ML Resource Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile applications based on AI/ML face challenges due to intensive computation, memory, and power consumption, leading to the need for offloading inference procedures to Internet data centers, with no existing methods for user equipment to effectively assist in resource management for AI/ML operations.
Innovation Solution
User equipment provides computation capability resource assistance information, including current capacity and expected consumption, to network devices through MAC control elements or RRC signaling, enabling optimized AI/ML operation allocation and resource management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If AI/ML inference procedures are offloaded to Internet data centers, then computation and power consumption on user equipment are reduced, but network dependency increases and latency may increase
Solution Approach 1:
The patent segments AI/ML operations into two parts: training on user equipment and inference on network devices. This allows the computationally intensive training to be performed locally using available data, while inference is offloaded to the network, resolving the contradiction between reducing local power consumption and maintaining response time.
Solution Approach 2:
The patent performs preliminary AI/ML model training on user equipment using local data before inference. This preliminary action enables the model to be pre-adapted to user-specific patterns, reducing the computational burden during inference while maintaining accuracy and reducing latency.
2Productivity
If more computation resources are allocated to AI/ML operations on user equipment, then AI/ML task performance improves, but energy consumption and computation cost increase
Solution Approach 1:
The patent divides AI/ML workloads into training (performed on user equipment) and inference (performed on network devices). This segmentation allows each component to be optimized independently, improving overall productivity while managing energy consumption by performing only lightweight inference locally.
Solution Approach 2:
The patent introduces a split scheduler as an intermediary that coordinates between user equipment and network devices. This mediator optimizes the distribution of computation tasks, ensuring that AI/ML operations are completed efficiently while balancing energy consumption across the system.
3Adaptability or versatility
If AI/ML operations are separated between user equipment and network, then resource utilization improves, but system complexity increases
Solution Approach 1:
The patent creates a universal framework where user equipment can perform both AI/ML training and traditional communication functions, while network devices handle inference. This multi-functional design improves resource utilization without significantly increasing system complexity by reusing existing communication infrastructure.
Solution Approach 2:
The patent implements feedback mechanisms where user equipment reports computation capability and data status to the network, and the split scheduler adjusts resource allocation accordingly. This feedback loop enables adaptive resource management that improves versatility while keeping system complexity manageable through automated control.
Data Source
AI summary
A reporting method includes: a user equipment (UE) providing computation capability resource assistance information of the UE to a network device. The computation capability resource assistance information may be carried in a MAC control element (MAC CE) or radio resource control (RRC) signaling, and the computation capability resource assistance information of the UE may be provided to the network device through a control channel carrying the MAC CE or the RRC signaling.