Research On Task Offloading Scheduling And Resource Provisioning Optimization Of Mobile Edge Computing Network

With the rapid development of the Internet of Things(IoT),device inter-connection and business volume has also shown explosive growth,which brings more requirements for the mobile edge computing(MEC)network to deal with massive increase in data.The MEC network,as a new IoT computing model,uses the mutual cooperation mechanism between edge nodes to transfer computing power from the cloud to the edge of the network to handle largescale and complex computing issues,thereby achieving network interconnection,collaboration,edge intelligent,safe offloading computing services.Compared with traditional cloud computing networks,MEC networks have the property of large scale,diversity and wide coverage.However,there are some new challenges in terms of computing coordination,resource optimization and security.Therefore,it has become the focus of industry and academic community to deal with these new challenges.In this context,this thesis investigates the theory of offloading scheduling and resource optimal allocation in MEC networks.The main research contributions are as follows.(1)Task offloading scheduling and resource provisioning optimization based on random online learning.To improve the quality of computation experience for mobile devices,MEC is a promising paradigm by providing computing capabilities in close proximity within a sliced RAN,which supports both traditional communication and MEC services.Nevertheless,the design of computation offloading strategies for a virtual MEC system remains challenging.This paper considers MEC for a representative mobile user in a sliced RAN,where multiple base stations(BSs)are available to be selected for computation offloading.The problem of solving an optimal computation offloading policy is modeled as a Markov decision process(MDP),where our objective is to maximize the long-term utility performance whereby an offloading decision is made based on the task queue state,the energy queue state as well as the channel qualities between mobile user and BSs.To avoid the high-dimensional state space,we decompose the MDP into a series of single-agent MDP with reduced state spaces,and derive an online localized learning algorithm to learn the state value functions.Numerical experiments show that our proposed learning algorithms achieve a significant improvement in computation offloading performance compared with the baseline policies.(2)Task offloading scheduling and resource provisioning optimization based on blockchain.Vehicle ad hoc networks(VANETs)play an irreplaceable role in intelligent transportation,but there are a lot of challenges in the deployment of in VANETs in terms of dealing with real-time,high-speed and continuous dataflow.Considering the requirements of real-time,high-speed and continuous data flow for communication between different physical entities in large-scale and dynamic mobility scenarios,this paper studies the allocation of computing resources and wireless bandwidth with offloading decision-making,consensus mechanisms,and joint optimization computations in order to avoid the inability to effectively adapt VANET's to handle complex practical problems in the dependency model and avoid illegal attacks in offloading.Furthermore,to realize the best offloading strategy for all vehicles in the VANETs,we introduce a novel deep reinforcement learning(DRL)based scheme.Experimental results show that the proposed method achieves better access control and offloading performance,and show a significant advantage over the existing solutions.(3)Coupling resource scheduling and resource provisioning optimization based on edge computing.Smart cities are the integration of cloud computing and cyber-physical systems(CPS).As the combination of CPS and cloud computing,sensor-cloud systems become popular in various fields.Therefore,the physical nodes sharing by multiple users can improve computing performance as well as bring conflicts.When a physical receives multiple service commands,we need couple resource management.In order to study the couple resource management problem,we propose an edge computing model that improves the Hungarian algorithm.The proposed model has shorter delay and is effective.In our model,the edge computing layer acts as buffer and controller between the CPS and the cloud layer.It can also deal with malicious attacks and ensure sustainability.Experimental results show that our method can manage couple resources well and achieve service sustainability.(4)Patrolling management scheduling and resource provisioning optimization based on strategy synthesis.The MEC systems are prone to attacks such as tampering and overhearing when offloading to edge servers.Therefore,we design an effective patrolling algorithm to protect the edge nodes in MEC.Of which,the patroller program is a software agent that can detect/prevent bad activities on the edge node.Considering the worst-case assumption that the actual attacker's ability is unknown,and a strong defense strategy is needed,we design an effective patrol strategy algorithm in the MEC environment.Based on this,a synthetic method is proposed to construct the defender strategy in the patroller game.We divide the game into disjoint subgames,and then solve the subgames recursively.The experimental results show that the proposed method is suitable for more general models.