Design And Optimization Of Low-Cost Edge Computing In Mobile Scenarios

As the era of the Internet of Things arrives,the number of Io T devices is constantly increasing.This proliferation of Io T devices has generated massive amounts of data and a variety of computing tasks,and cloud computing’s bandwidth and latency limitations make it difficult to meet such computing demands,leading to the emergence of edge computing.Simply put,edge computing moves the computing servers responsible for processing from the network center to the edge closer to the user,in order to reduce end-to-end latency and improve service quality.In recent years,the widespread adoption of mobile smart devices and their enhanced computing capabilities have made effective utilization of the ubiquitous computing power in the environment a hot topic in edge computing research.This thesis proposes a novel low-cost edge computing system for mobile scenarios,aimed at using mobile intelligent devices,primarily smartphones,in the environment as computing resources for the edge computing system,providing image detection-based computing services for front-end devices.The system obtains image data and computing tasks from front-end devices through deployed edge gateways and offloads tasks for computation to edge servers within the network coverage.Unlike traditional edge computing system solutions,this system does not require large-scale edge server deployment in advance,fully utilizing the idle ubiquitous computing power in the environment,reducing the cost and complexity of edge computing system deployment.In the system’s design process,this thesis proposes solutions to new problems brought about by the mobile scenario.To address the issue of unstable service due to mobility and wireless connections,the thesis proposes a task offloading algorithm for mobile scenarios based on latency restrictions,making task offloading decisions based on load information,mobility information,and network status information carried in the heartbeat package sent by the edge server.To address the problem of insufficient detection accuracy due to limited device computing power,the thesis proposes an optimized detection result strategy based on image segmentation,splitting images into multiple computing nodes for parallel computation,improving the target detection accuracy while ensuring low latency.To further optimize the service discovery process in the edge computing system,which may result in a large number of redundant connections,the thesis proposes a reverse service discovery mechanism.Subsequently,this thesis conducts a detailed requirements analysis and implementation of the proposed system,followed by experimental verification.The experimental results demonstrate that the task offloading algorithm proposed in this thesis substantially reduces the probability of request sending failure and result return loss by an average of 5% compared with the average allocation strategy;the object detection optimization strategy using image segmentation and result fusion boosts the average accuracy by60%;the reverse service discovery eradicates 98% of unnecessary connections.This work still has limitations,such as the lack of generality of the method to enhance the detection accuracy and the potential for improvement of the system performance.Future work will be devoted to addressing these issues and further advancing the performance and reliability of the edge computing system.