Vehicle Offloading Server Selection Using Mean-Field Evolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inefficient server selection and resource allocation in dense vehicle and server areas lead to increased network latency and computation times, particularly during peak hours, and pose challenges in managing heterogeneous computing tasks from vehicles like automated and emergency vehicles.
Innovation Solution
An optimization system using a mean-field evolution (MFE) model to estimate agent behavior and allocate resources efficiently, pre-sorting offloading requests based on priority and resource availability, and employing a probability distribution to rapidly select servers and allocate resources independently of area density or agent quantities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional server assignment and resource allocation schemes are used in dense vehicle and server areas, then system complexity is manageable, but network latency and computation times increase significantly
Solution Approach 1:
The patent transforms the discrete server selection problem into a continuous optimization problem by using probability distributions to represent server selection strategies. This parameter transformation enables the use of mean-field evolution to optimize selection parameters, achieving lower latency while maintaining manageable complexity through mathematical abstraction of the selection process.
Solution Approach 2:
The patent introduces mean-field evolution as an intermediary optimization layer between traditional server assignment and resource allocation. This intermediary mechanism uses probability distributions and evolutionary algorithms to mediate the selection process, reducing direct complexity while optimizing latency through iterative parameter refinement based on system characteristics.
2Productivity
If more servers are deployed to handle increased offloading demands during peak hours, then service capacity increases, but server selection and resource allocation times increase
Solution Approach 1:
The patent performs preliminary characterization of servers and vehicles, storing key characteristics in advance. This preliminary action enables rapid matching during peak hours without requiring real-time analysis of all system parameters, thus maintaining fast selection times even as service capacity scales with additional servers.
Solution Approach 2:
The patent transforms the scaling problem by using probability distributions that can be efficiently updated and sampled regardless of system size. This parameter transformation allows the selection mechanism to remain computationally efficient even as the number of servers increases during peak demand periods.
3Reliability
If resource allocation is optimized for automated vehicles, then automated driving performance improves, but emergency vehicles may experience critical delays
Solution Approach 1:
The patent implements dynamic resource allocation where the probability distribution for server selection is continuously updated based on real-time system characteristics and task priorities. This dynamic adjustment allows the system to prioritize emergency vehicles when detected, ensuring critical response times are met while maintaining optimized performance for automated vehicles during normal operations.
4Ease of manufacture
If conventional resource allocation methods are used, then implementation is straightforward, but efficient server selection becomes difficult in dense areas with numerous servers and heterogeneous vehicles
Solution Approach 1:
The patent changes the fundamental parameters of server selection from discrete, combinatorial choices to continuous probability distributions. This parameter transformation simplifies the implementation by enabling the use of efficient sampling methods while dramatically improving selection efficiency in dense environments with numerous servers and heterogeneous vehicle types.
Data Source
AI summary
System, methods, and other embodiments described herein relate to selecting servers and allocating resources concurrently for offloading computing tasks from vehicles. In one embodiment, a method includes acquiring characteristics of a vehicle and a server for an offloading request, wherein the offloading request is associated with a computing task of the vehicle. The method also includes, upon satisfying criteria for optimization associated with executing the computing task remotely, determining server selection and resource allocation by processing the characteristics using modeling. The method also includes communicating the server selection and the resource allocation to the vehicle.