Research On Task Offloading For Edge Computing In Heterogeneous Network

With the rapid development of internet of things and mobile communication technology,the smart mobile terminals and computing intensive services(such as smart hourse,automatic driving,smart city,etc.)have been increasing sharply,which driving the explosive growth of emerging applications and network traffic.In the future mobile communication technology,the Heterogeneous Network(Het Net)has become an inevitable development trend,and the convergence of diverse networks will become a normal.With the heterogeneity of network and the cloudization of mobile computing,Mobile Edge Computing(MEC)technology for heterogeneous networks has become a hot research topic.Under the constraints of limited network edge resources(computing and storage resources,etc.),the heterogeneity of network has an important impact on the decisionmaking of MEC task offloading and resource allocation strategy.In order to solve the problem of MEC task offloading and computing resource allocation in heterogeneous network environment,a task offloading queue model is established based on lyapunov optimization method.By analyzing the tradeoff between task offloading utility and offloading delay,a joint optimization of the task offloading and computing resource allocation is proposed.In addition,in order to get the optimal offloading strategy,a heuristic task offloading algorithm based on a search tree is proposed for heterogeneous network in edge computing.Meanwhile,in order to reduce the complexity of the algorithm,a offloading criterion based on priority is designed.The simulation results show that the proposed algorithm can reduce the complexity of the algorithm and improve the efficiency of task offloading while making a tradeoff between the task offloading utility and delay in heterogeneous network environment.Considering the dynamic characteristics of the channel quality of communication link and task request information,The influence of this dynamic characteristics on task offloading decision is studied in this thesis.The task offloading problem is formulated as a Markov Decision Process(MDP)problem.In order to minimize the long-term average cost of task offloading,a task offloading strategy based on Q-learning is proposed.Then,in order to overcome the high-dimensional problem of state space and accelerate the learning speed,a task offloading strategy based on DQN is proposed.Using DQN's self-learning strategy to get the optimal decision of task offloading and compuation resource allocation.Finally,simulation results show the effectiveness of the proposed strategy.