User Equipment ML Location Policies for Dynamic Task Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing telecommunications networks face challenges in managing artificial intelligence/machine-learning tasks due to limited data transmission and energy capacities, necessitating a higher degree of flexibility and controllability in task distribution between user equipment and network entities.
Innovation Solution
A method for transmitting and using user equipment machine-learning location selection policy information, enabling interaction between user equipment and a machine-learning server entity through a machine-learning interface, allowing flexible adaptation of AI/ML tasks based on network policies and capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI/ML tasks are performed on user equipment, then application responsiveness and user privacy are improved, but processing resources, memory, and energy consumption increase
Solution Approach 1:
The patent segments AI/ML tasks into multiple components that can be distributed between user equipment and network entities. The policy information divides task execution locations into different segments (UE-side, edge-side, core-side), allowing the system to select appropriate segmentation based on energy constraints and performance requirements.
Solution Approach 2:
The patent implements dynamic task distribution through policy information that can be updated and adjusted based on current network conditions, device state, and energy availability. The machine-learning agent dynamically selects execution locations rather than using a static allocation, enabling adaptation to changing energy constraints.
2Use of energy by moving object
If more AI/ML tasks are offloaded to network entities, then user equipment energy consumption is reduced, but data transmission requirements and network complexity increase
Solution Approach 1:
The patent introduces policy information as an intermediary layer between user equipment and network entities. This policy information acts as a mediator that guides the machine-learning agent in selecting appropriate task execution locations, simplifying the decision-making process and reducing network complexity through structured policy rules.
Solution Approach 2:
The user equipment's machine-learning agent autonomously selects task execution locations based on received policy information, without requiring complex centralized control. The agent serves itself by making local decisions based on policy guidelines, reducing the need for complex network-wide coordination mechanisms.
3Adaptability or versatility
If flexible task distribution is implemented, then adaptability to different scenarios is improved, but control mechanisms and policy management complexity increase
Solution Approach 1:
The patent uses policy information containing adjustable parameters (such as energy thresholds, latency requirements, and task priorities) to control task distribution behavior. By changing these parameters, the system adapts to different scenarios without requiring fundamental changes to the control mechanism, maintaining simplicity while achieving flexibility.
Data Source
Figure 1~3
Figure 4
AI summary
The invention relates to a method for transmitting and/or using a user equipment machine-learning location selection policy information when operating a user equipment, or when operating a user equipment connected to a telecommunications network, wherein the user equipment comprises a user equipment machine-learning agent, the user equipment machine-learning agent being configured to interact with a machine-learning server entity in order to perform machine-learning operations involving the machine-learning server entity solely or involving the user equipment solely or involving both the machine-learning server entity and the user equipment, wherein, in order to transmit and/or in order to use a specific user equipment machine-learning location selection policy information, the method comprises the following steps: -- in a first step, the user equipment comprises or receives the specific user equipment machine-learning location selection policy information to be evaluated and/or to be applied by the user equipment machine-learning agent, -- in a second step, the user equipment machine-learning agent and the machine-learning server entity interact in accordance with a machine-learning interface for control plane data and user plane data.