Research On Application Sets Data Pricing Based Task Offloading In Multi-access Edge Computing

Multi-access Edge Computing(MEC)reduces latency and energy consumption of task execution by offloading computing tasks from End Nodes(ENs)to MEC servers.Applying data pricing strategies in task offloading is perceived as a promising solution for Internet Service Providers(ISPs)to cope with the ensuing network congestion.However,traditional application usage pattern aware task offloading strategies do not consider the sequence patterns among the applications used by users,which results in lower ISP’s profit and user satisfaction.In this thesis,a novel scheme named Associated-app Usage Patterns Aware task offloading scheme(AUPA)is proposed to alleviate the tension between traffic data supply and user satisfaction.First,a task offloading dataset is built based on the real application usage dataset.This thesis treats applications as the smallest unit in the task offloading process and mines the sequential patterns among the associated applications to extract temporal association rules.Then,taking partial offloading and capacity constraints into account,a smart data pricing strategy is designed based on these association rules to regulate task offloading decisions.Finally,the ISP’s profit maximization problem is formulated as a Nonlinear Programming(NLP)problem based on partial offloading and solved using the Stackelberg game theory,while ensuring MEC capacity constraints.The performance of proposed solution is evaluated in terms of ISP’s profit,consumers’ surplus,capacity utilization,and traffic efficiency.Experimental results show that the task offloading scheme based on associated application data pricing achieves better ISP’s profit with guaranteed capacity constraints than other baseline schemes,especially in peak hours while there are many types of applications and frequent switching between applications,proving that the scheme proposed in this thesis is more suitable for task offloading scenarios in MEC systems that need to consider user characteristics.