Cellular AI/ML Task Assignment for Low-Latency UE Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile communication systems lack methods to efficiently support AI/ML services, particularly in terms of latency minimization, requiring network assistance for task assignment and resource management.
Innovation Solution
A method and apparatus for AI/ML job execution involving UE and ASP servers, utilizing a cellular network to determine optimal task combinations based on expected completion time and available resources, adjusting policies and PDU sessions for low latency performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If AI/ML tasks are executed on user equipment without network assistance, then device autonomy is maintained, but latency cannot be minimized and resource optimization is limited
Solution Approach 1:
The network function acts as an intermediary between the user equipment and the task execution environment. It receives task information from the UE, determines optimal execution resources by analyzing device capabilities and network conditions, and coordinates task assignment. This mediator approach enables latency minimization through centralized intelligence while keeping the UE implementation relatively simple.
Solution Approach 2:
The network function performs preliminary analysis of device capabilities, resource availability, and task requirements before task execution begins. By pre-determining the optimal execution plan and resource allocation, the system minimizes latency during actual task execution without requiring complex real-time decision-making at the device level.
2Productivity
If more tasks are assigned to UE to improve resource utilization, then system productivity increases, but device resource constraints may be exceeded
Solution Approach 1:
The network function dynamically adjusts task assignment based on real-time device resource availability and capability information. It continuously monitors device status, network conditions, and task requirements to optimize the distribution of tasks across multiple devices, maximizing system productivity while ensuring no single device is overloaded beyond its resource constraints.
Solution Approach 2:
Complex AI/ML tasks are segmented into smaller sub-tasks that can be distributed across multiple user equipments. The network function divides the overall task workload and assigns appropriate segments to different devices based on their specific resource availability and capabilities, thereby increasing system-level productivity while maintaining reliability at the individual device level.
3Loss of time
If task assignment is optimized for latency, then service quality improves, but system complexity increases
Solution Approach 1:
The complex task assignment and optimization logic is extracted from the user equipment and centralized in the network function. The UE simply submits task requests and receives assignments, while the network function handles the computationally intensive optimization of task distribution, resource allocation, and latency minimization. This extraction reduces device complexity while maintaining optimized task completion.
Data Source
AI summary
A method and apparatus for performing AI/ML job through the steps of: receiving expected completion time and available resources of UE for a plurality of candidate combinations of a plurality of tasks included in a job from the UE via a cellular network; determining a candidate combination from among the plurality of candidate combinations based on the expected completion time and available resources of the UE for the plurality of candidate combinations; and assigning a task according to determined candidate combination to the UE through a PDU session of the cellular network.