Decentralized Resource Allocation for Edge Computing Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently managing radio and computing resources owned by different entities, such as mobile network operators (MNOs) and computing resource providers (CRPs), particularly in minimizing latency and addressing strong coupling between providers in co-located mobile edge computing environments.
Innovation Solution
Formulating a resource allocation problem as a generalized Nash equilibrium problem (GNEP) and converting it into a Nash equilibrium problem (NEP), with resource allocation occurring on a penalty basis, to manage radio and computing resources effectively between MNOs and CRPs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If resource allocation is managed separately by MNO and CRP without coordination, then each entity can independently optimize its own resources, but the overall system latency increases and resource utilization efficiency deteriorates due to strong coupling between providers
Solution Approach 1:
The patent merges the resource allocation decisions of MNO and CRP into a unified GNEP framework. Instead of separate independent optimizations, both entities' resource allocation variables (radio resources w and computing resources m) are combined into a joint optimization problem with coupled constraints, allowing simultaneous optimization of both resources to minimize overall system latency.
Solution Approach 2:
The patent introduces penalty parameters (κ) as intermediaries to handle the coupling constraints between MNO and CRP resource allocation. These penalty parameters act as mediators that enforce the coupling constraints (g(w,m)≤0) during the iterative optimization process, enabling coordinated resource allocation while maintaining the decentralized decision-making structure of the two entities.
2Productivity
If a GNEP-based algorithm is used to solve the resource allocation problem, then resource allocation efficiency improves, but the computational complexity and difficulty of solving the problem increases
Solution Approach 1:
The patent segments the complex GNEP solution process into iterative steps using penalty-based algorithms. The original difficult GNEP is broken down into a sequence of simpler constrained optimization problems that can be solved iteratively. Each iteration solves a sub-problem with updated penalty parameters, making the overall complex problem tractable through stepwise decomposition.
Solution Approach 2:
The patent applies dynamic penalty parameters (κp,k) that evolve during the iterative optimization process. The penalty parameters are updated dynamically based on the violation of coupling constraints in each iteration, allowing the algorithm to adaptively balance between satisfying constraints and optimizing the objective function, thereby managing computational complexity dynamically.
3Manufacturing precision
If coupling constraints between MNO and CRP are enforced, then resource allocation accuracy improves, but the flexibility and ease of operation decreases
Solution Approach 1:
The patent uses dynamic penalty parameters that adjust during iterations to enforce coupling constraints. The penalty parameters start at initial values and are updated based on constraint violations, providing a flexible path from initial resource allocations to final allocations that satisfy coupling constraints, thereby maintaining operational flexibility while achieving accuracy.
Solution Approach 2:
The patent implements feedback through the penalty mechanism where constraint violations are detected and fed back into the optimization process via updated penalty parameters. This feedback loop continuously adjusts the resource allocation decisions to satisfy coupling constraints, ensuring accuracy while maintaining the ability to adapt to changing conditions through iterative refinement.
Data Source
AI summary
The present disclosure relates to a technical idea for managing radio and computing resources in coexistence edge computing. A method of allocating radio and computing resources in coexistence edge computing according to one embodiment may include a step of formulating a resource allocation problem for two different entities with conflicting relationships in minimizing latency as a generalized Nash equilibrium problem (GNEP), a step of converting the formulated GNEP into a Nash equilibrium problem (NEP), and a step of allocating resources on a penalty basis for the converted NEP.