Task scheduling method based on edge computing

Abstract

The invention relates to a task scheduling method based on edge computing, which comprises the following steps that: a strategy network is maintained in each region, and each region independently schedules tasks received in real time in the region based on the strategy network and a real-time environment; network parameters of the strategy network are updated online by adopting a deep reinforcement learning algorithm based on federated learning, specifically, historical data are stored in each region to form a local data set, deep reinforcement learning training is carried out based on the local data set, the network parameters are updated, and the training target is to minimize reward values of all tasks; and each region sends own network parameters to the central base station, and the central base station feeds back the updated network parameters to each region after performing unified updating based on federal learning. Compared with the prior art, the method has the advantages of enhancing model expansibility, protecting data privacy, improving system performance and the like.

Description

technical field [0001] The invention relates to the technical field of edge computing, in particular to a task scheduling method based on edge computing. Background technique [0002] Large-scale connectivity is one of the most challenging requirements for IoT networks, requiring efficient, scalable, low-complexity and privacy-friendly management of network resources. In addition, due to the limited computing and storage resources of IoT devices, a large number of resource-intensive tasks cannot usually be processed within the expected time, and uploading tasks to the cloud for processing may result in unbearably high latency. Therefore, edge computing is considered to be a promising technology because it can place computing and caching services at the edge of the network closer to users for processing. Edge computing is flexible and efficient. When scheduling, it usually takes minimizing cost and minimizing delay as scheduling goals, and completes scheduling with the optim...

AI Technical Summary

Problems solved by technology

[0004] However, the existing scheduling algorithm will bring a series of problems, mainly including the locality and inaccuracy of scheduling, and often fall into the situation of local optimal solution due to ignoring historical information

Applying traditional deep learning technology directly to task scheduling in distributed systems will also lead to loss and inaccuracy of environmental information, and over-fitting problems in the training process lead to over-reliance on training data

The method of directly using reinforcement learning for task scheduling not only lacks scalability, but also poses challenges to resource consumption and data privacy during massive data transmission.

Images