Research On Dynamic Computation Offloading Scheme In Time-varying Environment Of Ultra Dense Network

With the development of mobile communication technology and the widespread popularity of various mobile smart terminals,a large number of emerging mobile Internet services such as high-definition video transmission,augmented reality/virtual reality,and autonomous driving have emerged one after another.In this context,in order to further meet the delay and energy consumption requirements of many applications on the network,the industry has proposed the concepts of mobile edge computing and ultra-dense networks.Based on both,network carrying capacity and network service capacity have been effectively improved.On the one hand,the converged architecture based on the two can effectively extend computing and storage capabilities to the edge of the network;on the other hand,the ultra-dense networks technology also provides a solution to the connection problems caused by the expansion of user equipment.In addition,computation offloading can also effectively guarantee the delay and other requirements of end users in the network.Although there has been a lot of related work,the existing converged architecture design and computation offloading schemes are not fully considered.How to effectively combine the two for architecture design and the corresponding deployment plan can be further studied.Besides,the existing computation offloading schemes seldom consider the dynamic changes of the environment and the different needs of user services,and there is still room for further research.Therefore,this subject has conducted in-depth research on the design of the convergence architecture of mobile edge computing and ultra-dense networks.Taking into account the characteristics of ultra-dense networks and mobile edge computing technology and the needs of network services,the necessary design criteria for the network architecture is proposed.And also this thesis designed specific hierarchical functional modules.Based on the container and Kubernetes,this thesis proposed the corresponding deployment plan.In addition,for the computation offloading scheme,this topic focuses on multi-user computation offloading in dynamic time-varying scenarios,taking into account the dynamic changes of channels and user tasks.And for the multi-user computation offloading problem,considering the user's competitive relationship,the overall model is transformed into a stochastic game problem,and the overall delay and packet loss of the user computing task are optimized,and finally solved based on the Nash-Q learning algorithm.The simulation part verifies the performance of the scheme proposed in this topic.Based on the comparison with the numerical simulation of common schemes,the scheme in this thesis can deal with related problems more effectively.