Task Offloading in Mobile Edge Computing via Stochastic Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In mobile edge computing (MEC) networks, the uncertainty of task queue waiting times at MEC servers leads to unpredictable computing delays and high energy consumption, as existing methods fail to accurately account for queuing delays and focus primarily on user-side energy consumption, neglecting the limitations of MEC servers with limited resources.
Innovation Solution
A task offloading and resource allocation method using two-stage stochastic programming to model queue waiting times as random parameters, transforming the problem into a sample mean approximation, and decoupling it into sub-problems for local computing resource allocation, transmission power, and edge computing resource allocation, using Lagrange multipliers and genetic algorithms to optimize resource allocation strategies under delay constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional task offloading methods are used that ignore queue waiting time, then user-side energy consumption is optimized, but computing delay becomes uncertain and unreliable
Solution Approach 1:
The patent applies preliminary action by predicting queue waiting time before task offloading decisions are made. The system performs preliminary estimation of the task queue waiting time at the MEC server based on historical data and current system state, then uses this prediction to inform the offloading decision. This allows the system to account for future queuing delays without waiting for actual queue conditions, thereby improving computing delay reliability while maintaining energy optimization.
2Loss of time
If MEC servers allocate more computing resources to handle burst requests, then queue waiting time decreases, but system energy consumption increases
Solution Approach 1:
The patent applies dynamics by implementing dynamic resource allocation at the MEC server. Instead of allocating fixed computing resources, the system dynamically adjusts the computing frequency and resource allocation based on real-time queue conditions, task characteristics, and energy constraints. The MEC server can scale resources up during burst periods to reduce waiting time, then scale down during low-load periods to conserve energy, achieving a balance between responsiveness and energy efficiency.
Solution Approach 2:
The patent applies parameter changes by optimizing multiple parameters simultaneously including computing frequency, transmission power, and resource allocation ratios. The system changes these parameters dynamically based on the predicted queue waiting time and current system state, using optimization algorithms to find the optimal parameter combination that minimizes both queue waiting time and energy consumption rather than fixing any single parameter.
3Manufacturing precision
If accurate prediction of task queue waiting time is attempted, then computing delay can be optimized, but system complexity increases significantly
Solution Approach 1:
The patent applies this principle by using lightweight, computationally efficient prediction models that can be quickly updated and discarded. Instead of employing complex, resource-intensive prediction algorithms, the system uses simplified models that provide sufficient accuracy for offloading decisions while maintaining low computational overhead. These prediction models can be rapidly recalculated based on changing system conditions without requiring significant processing resources.
Solution Approach 2:
The patent applies partial action by implementing prediction and optimization only for the most critical parameters and scenarios. Rather than attempting to predict and optimize all possible variables in the system, the focus is placed on the key factors that most significantly impact computing delay and energy consumption, such as queue waiting time and computing frequency. This selective approach achieves good performance while keeping system complexity manageable.
Data Source
AI summary
A task offloading and resource allocation method in an uncertain network environment is provided. A task offloading process is modeled as a two-stage offloading model. The model is optimized to a task offloading and resource allocation problem based on two-stage stochastic programming. Based on a stochastic simulation algorithm, the task offloading and resource allocation problem is transformed to a sample mean approximation problem. The sample mean approximation problem is decoupled to a local computing resource allocation sub-problem, a transmission power and edge computing resource joint allocation sub-problem, and an offloading decision sub-problem. The three sub-problems are solved respectively by using a standard Lagrange multiplier algorithm, by using a genetic algorithm, and by analyzing delay estimation and energy consumption budget of local computing and delay estimation and energy consumption budget of edge computing. The user performs task offloading based on an optimal allocation strategy obtained by solving the three sub-problems.