Research On Dynamic Perception Scheduling Strategy Of Edge Intelligent Computing Based On MQTT

In recent years,with the development of deep learning lightweight algorithms for face recognition and AI hardware acceleration,traditional devices also have the ability of intelligent video analysis.This paper focuses on the communication field of edge intelligent computing.It fully considers the issues of large differences in the network environment of the intelligent device at the edge and vulnerability to attacks.It focuses on the edge intelligent computing queue scheduling algorithm and communication protocol related algorithm,and completes the design and implementation of the embedded real-time communication system.The research in this paper will contribute to the reliability and security of the real-time communication scheduling system of edge intelligent computing,and it has extremely important theoretical and practical value.The main work of this paper is as follows:According to the requirements of edge intelligent computing Qo S,a dynamic sensing queue scheduling algorithm based on WF~2Q+is researched and designed.Because the traditional WF~2Q+has a fixed weight for each data stream,real-time data cannot be transmitted preferentially when the scheduling rate is low.This paper proposes a comprehensive PQ algorithm to ensure the highest priority of core real-time data.The improved WF~2Q+algorithm dynamically adjusts the weight of non-core real-time data streams according to the length of the queue backlog.Therefore,the improved WF~2Q+combined with PQ algorithm can not only satisfy the real-time data transmission,but also have certain fairness.In addition,a large amount of data suddenly sent to the network may cause network congestion.Token bucket technology is used to shape the traffic.According to the characteristics of limited intelligent software and hardware resources at the edge and vulnerability to attacks,a secure and reliable data transmission method based on MQTT was researched and designed.Because MQTT does not use encryption for communication by default,the risk of data eavesdropping,tampering and impersonation.This article proposes using SSL/TLS security authentication scheme at the network transport layer,and using a username and password for authentication at the application layer.For the topic access control mechanism in the message queue,an ACL table is used to constrain it.Due to the limited software and hardware resources,large files are transmitted in sub-contracting mode,while mechanisms such as stop waiting,sliding windows,and timeout retransmissions are used to ensure reliable data transmission.Since MQTT is a lightweight protocol suitable for M2M communication,it is not suitable for distributed storage of massive Io T data.This paper proposes to send the data received by the MQTT protocol to the kafka message queue.Finally,based on the WF~2Q+based dynamic sensing queue scheduling algorithm and MQTT-based secure and reliable transmission method,this paper proposes a real-time communication system solution for edge intelligent computing,and builds a real-time communication system based on edge intelligent computing.The experimental results show that the communication system designed in this paper can meet the needs of edge intelligent computing application scenarios in terms of real-time,security and reliability.