| Next-generation wireless networks must support ultra-reliable,low-latency communication and intelligently manage a massive number of Internet of Things(IoT)devices in real-time,within a highly dynamic environment.This need for stringent communication quality-of-service(QoS)requirements as well as mo-bile edge and core intelligence can only be realized by integrating fundamental notions of artificial intelligence(AI)and machine learning across the wireless infrastructure and end-user devices.In this context,it is necessary to investigate the application of the machine learning and artificial intelligence techniques for wireless communications for controlling and managing the wireless networks in a self-organizing,smart,and real time way.Recently,machine learning algorithms can be used for wireless networks in two ways.First,the machine learning algorithms can be used to identify the wireless networking environment and control and manage the wireless network-ing resource and users according to the wireless networking environment.Sec-ond,the machine learning algorithms can be used to predict the network states and mobile users behaviors so as to improve the network performance.In this thesis,the research mainly focuses on how to use machine learning algorithms to manage and schedule the spectrum resource allocation and user association,how to use machine learning algorithms to predict the network states and users behaviors,and how to design novel machine learning algorithms for the opti-mization of the wireless networks.In particular,this thesis first introduces the architecture of novel wireless networks and its corresponding challenges and opportunities.Based on the proposed architecture,we introduce how to use machine learning algorithms to manage the spectrum resource over licensed and unlicensed band,cache,the deployment of UAVs,and 360°content deliv-ery and caching.The key contribution is given as follows:(1)Study of the resource allocation over licensed and unlicensed bands for uplink and down-link decoupling users.The echo state network(ESN)based machine learning algorithm is proposed to optimize the spectrum allocation and user association so as to maximize the users’ data rates.(2)Investigation of the cache manage-ment.The use of the users’ behavior information is proposed to determine the cached contents so as to reduce the traffic load over fronthaul and backhaul and maximize the users’ effective capacities.The ESN-based prediction algorithm is proposed to predict the users’ mobility patterns and content request distri-butions so as to determine the cached contents at baseband units and remote radio heads.(3)Optimization of the deployment of UAVs,the use of a concep-tor based ESN algorithm is proposed to analyze and predict the content request distribution and mobility patterns of each user.According to the predictions,one can determine the optimal user association,UAV locations,and cached con-tents so as to maximize the quality of experience of each user.(4)Management of 3600°content delivery and caching.To overcome the disadvantage of liquid state machine,joint liquid state machine and ESN based machine learning al-gorithm is proposed for the optimization of 3600°content delivery and caching so as to maximize the users’reliability. |
