Edge-End Collaborative Scheduling for Heterogeneous Tasks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for scheduling heterogeneous tasks in multi-access edge computing environments struggle with resource fragmentation, interference, and inaccurate resource estimation, leading to increased processing delays and reduced quality of service.
Innovation Solution
A digital twin-based edge-end collaborative scheduling method using multi-agent deep reinforcement learning, which virtualizes and models heterogeneous computation and communication resources, optimizes task offloading, and allocates resources to minimize processing delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If multi-access edge computing is used to process tasks at edge servers, then task processing delay is reduced, but communication resource competition and fragmentation are aggravated
Solution Approach 1:
The patent segments the resource scheduling problem into multiple independent sub-problems by introducing separate scheduling modules for computation resources and communication resources. The computation resource scheduler handles task offloading decisions while the communication resource scheduler manages bandwidth allocation, allowing each module to optimize its specific resource type without interfering with others, thus resolving the contradiction between reducing processing delay and maintaining communication resource efficiency
Solution Approach 2:
The patent introduces a digital twin as an intermediary virtual model that simulates the edge computing environment and predicts resource availability, transmission quality, and processing delays. This intermediary enables the scheduling system to make informed decisions about resource allocation by evaluating multiple scenarios before execution, thereby reducing actual processing delay while avoiding communication resource fragmentation through proactive planning
2Productivity
If heterogeneous tasks perform highly concurrent access, then system throughput increases, but transmission conflicts and QoS reduction occur
Solution Approach 1:
The patent applies preliminary action by using the digital twin to predict future resource availability and transmission quality in advance. The system pre-schedules tasks and allocates communication resources based on predicted conditions, allowing heterogeneous tasks to be coordinated before actual execution. This prevents transmission conflicts from occurring in the first place while maintaining high system throughput through efficient resource utilization
Solution Approach 2:
The patent implements dynamic scheduling that adapts to changing network conditions in real-time. The system continuously updates its resource allocation decisions based on actual measurements from the physical environment compared against the digital twin model, allowing it to dynamically adjust to varying task requirements and network states. This dynamic approach enables the system to maintain both high throughput and transmission quality by responding to instantaneous conditions
3Device complexity
If traditional scheduling methods are used in dynamic network environments, then implementation simplicity is maintained, but state space explosion makes them ineffective
Solution Approach 1:
The patent creates a digital twin copy of the physical edge computing environment that replicates its state, resources, and dynamics. Instead of trying to manage the complex real-time scheduling problem directly in the physical system, the patent uses the digital twin as a simplified virtual replica where scheduling decisions can be simulated and optimized. This copying approach transforms the intractable real-time scheduling problem into a manageable virtual modeling problem, resolving the contradiction between algorithm simplicity and processing efficiency
Data Source
AI summary
A digital twin-based edge-end collaborative scheduling method for heterogeneous tasks and resources includes the following steps: establishing an edge wireless network based on digital twin; constructing an edge-end collaborative scheduling problem prototype of the heterogeneous tasks and resources; performing problem conversion based on a multi-agent Markov decision process; constructing an Actor-Critic neural network model based on multi-agent deep reinforcement learning; performing offline centralized training of the neural network model by digital twin; performing online distributed execution of task offloading and computation and communication resource allocation by end devices to collaboratively process the heterogeneous tasks. The method optimizes the heterogeneous computation resource types, the task offloading ratio, the transmit power of the end devices and the computation resource allocation ratio of edge servers through digital twin, supports the on-demand offloading of heterogeneous tasks, realizes edge-end collaborative computing, and minimizes the total task processing delay.