In recent years,due to the rapid development of the mobile Internet,the amount of data generated by devices located at the edge of the network has also begun to increase sharply,and how to deal with the massive data generated by the edge has also brought challenges to people.Although cloud computing,as the core technology of the Internet,can meet the requirements for processing massive amounts of data,it also has some shortcomings: it occupies a large amount of network bandwidth and the real-time requirements cannot be met.This gave birth to the concept of edge computing.Because edge computing delivers computing power to the edge of the network,it is far more effective than cloud computing in terms of real-time performance.At the same time,it does not need to transmit large amounts of data to the cloud to reduce the pressure on network bandwidth.This paper is mainly based on edge computing,and builds a robotic vision system to offload the computationally intensive target detection application to the edge server deployed at the edge of the network,thereby reducing the delay caused by model inference.The platform is mainly divided into three parts: terminal part,edge server part and cloud server part.The terminal is mainly responsible for controlling the robot through the ROS system,and it also undertakes part of the model calculation,which is mainly connected to the edge network through WIFI;the edge server is deployed at the edge of the network,such as the campus network intranet,responsible for the calculation of the offloading plan generation and Perform the remaining part of the model calculation;the cloud server is deployed on the public network and is responsible for some macro-level functions,such as: real-time monitoring of the robot’s status,verification of equipment,and application deployment.Unlike cloud computing,for cloud computing,applications only need to be deployed on the cloud server,while edge computing requires the same version of the application to be deployed on the terminal and the edge at the same time to achieve computation offloading.In addition,in order to ensure the reliability of edge-side services,a cluster was constructed using Kubernetes to avoid the unavailability of the entire edge service due to the failure of a certain working node.Through this system,the average time consumption of applications under different conditions can be improved,the energy consumption of robots performing calculations locally,and the problem of network congestion caused by a large number of image frames uploaded to the cloud can also be avoided. |