Adaptive Encapsulation and Cooperative Reasoning Method in Task Flow Scenarios with Delay Constraints

Abstract

The invention discloses an adaptive encapsulation and collaborative reasoning method in a time-delay-constrained task flow scenario, comprising the following steps: establishing an adaptive edge intelligent reasoning model, including terminal equipment for receiving tasks and making task offloading decisions the edge server; build the branch model of the edge server and the terminal device, the branch model includes several branches, each branch is equipped with a division point and an exit point, and the exit point is set at the terminal of each branch; according to the task The processing flow of the task is predicted respectively on the terminal device and the edge server, and the total processing time of the task; the objective function is set, if all the information of the task flow is known, the offline algorithm of dynamic programming is used to optimize the objective function, Otherwise, an online algorithm of deep reinforcement learning is used to optimize the objective function. The invention improves the number of tasks to be completed under the delay constraint and the accuracy of task inference.

Description

technical field [0001] The invention relates to the field of edge technology and deep learning technology, in particular to an adaptive encapsulation and collaborative reasoning method in a task flow scenario with time delay constraints. Background technique [0002] With the rapid development of the Internet of Things, more and more devices are connected to the Internet. Therefore, massive amounts of data are generated on terminal devices. Due to the limitation of network bandwidth, the traditional method of deploying artificial intelligence (AI) algorithms on cloud servers has been unable to meet many applications with delay-sensitive characteristics. At the same time, research on the scope and level of model encapsulation in many fields at home and abroad is relatively extensive. From the neural network itself to service calls to functional containers, the research and implementation of different scopes are relatively mature. However, there are still relatively large res...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the technical problem that the existing technology does not consider the interaction of adjacent tasks in the task flow to realize adaptive collaborative reasoning, the present invention proposes an adaptive encapsulation and collaborative reasoning method in the context of task flow with delay constraints, which solves the problem of existing The technology is based on experimental evaluation to achieve adaptive collaborative reasoning, and rarely achieves system optimization through theoretical modeling

Images